tag:blogger.com,1999:blog-43235994903429565322024-02-07T14:11:15.577-05:00Competitive Advantage via Quantitative MethodsA clearing house for examples, technologies, and theories involving the use of quantitative methods to create competitive advantage.CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.comBlogger37125tag:blogger.com,1999:blog-4323599490342956532.post-25633859856513661292013-12-29T17:28:00.001-05:002013-12-29T17:28:51.272-05:00Zara, Fast-fashion, Optimization and the OM Triangle<div dir="ltr" style="text-align: left;" trbidi="on">
<a href="http://www.zara.com/">Zara</a> is the dominant brand in the <a href="http://www.inditex.com/en">Inditex</a> apparel portfolio, which Inditex has operated in ‘<a href="http://en.wikipedia.org/wiki/Fast_fashion">fast-fashion</a>’ style for a few decades. Although other fast-fashion retailers such as <a href="http://www.hm.com/us/">H&M</a> and <a href="http://www.benetton.com/us/">Benetton</a> have been successful, none have achieved Zara’s level of success. Driven by extraordinary revenue growth rates, Zara has overtaken <a href="http://www.gap.com/">Gap</a> as the largest apparel retailer in the world and continues to expand into new geographies and new product lines such as intimate apparel (<a href="http://www.oysho.com/">Oysho</a>), home furnishings (<a href="http://www.zarahome.com/webapp/wcs/stores/servlet/home/zarahomeus/-15/80279977">Zara Home</a>) and apparel for young women (<a href="http://www.stradivarius.com/">Stradivarius</a>).<br />
<br />
The fast-fashion retail model uses drastically shortened production cycles to get apparel from concept to customer in weeks versus the month-long design and production cycles. This improvement in inventory turnover gets customers to return more frequently, <i>“According to an online poll, the constant newness in Zara's stores means that customers visit an average of 17 times a year, up from five or fewer times for rival shops.”</i> (Hall) (Ghemawat) But <i>“…Zara can lay a strong claim to owning the most impressive manufacturing and distribution process in the apparel industry.”</i> (IT)<br />
<br />
Perhaps the greatest contradiction though, is that Zara’s culture favors <i>“human intuition, vision and judgment (as opposed to analytical methods) for decision‐making purposes”</i> (Caro) so their more recent collaborations with <a href="http://www.mit.edu/~orc/">MIT’s Operations Research</a> department were labeled as groundbreaking. While the 77 year old founder of Zara, Amancio Ortega Gaona, never completed high school he brought in Jose Maria Castellano, <i>“who had a doctorate in business economics and professional experience in information technology, sales, and finance. In 1985, Castellano joined Inditex as the deputy chairman of its board of directors…”</i> (Ghemawat) Castellano developed all of the logistics systems during the 80’s that paved the way for rapid growth in the 90’s. (Hall) It is therefore logical that despite the seeming non-quantitative nature of the apparel business that some groundbreaking practices would emerge given the background of its deputy chairman. Some of the most noteworthy practices are cited next.<br />
<br />
<br />
<b>Quantitative Methods</b><br />
<br />
<div class="separator" style="clear: both; text-align: center;">
<a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEipVkOwv7cpE2LeFEsGxKN4CVjtQzjeWRZJDY9zmutp7zJlU9ddmi0fmO92qPQ96QJulaHr-WZObrIP67lU0Tin5y_hNCgAKsMDaOQ6eszH09mXPu6pcfQ1MpLY896Bd1ERNlLnuFdsWxOc/s1600/Zara+OM+Triangle.JPG" imageanchor="1" style="clear: left; float: left; margin-bottom: 1em; margin-right: 1em;"><img border="0" height="640" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEipVkOwv7cpE2LeFEsGxKN4CVjtQzjeWRZJDY9zmutp7zJlU9ddmi0fmO92qPQ96QJulaHr-WZObrIP67lU0Tin5y_hNCgAKsMDaOQ6eszH09mXPu6pcfQ1MpLY896Bd1ERNlLnuFdsWxOc/s640/Zara+OM+Triangle.JPG" width="369" /></a></div>
Utilization vs. Wait Times: Probably the greatest operational insight aiding Zara’s success, is that there is an inverse relationship between utilization (i.e. warehouses, production, or factories) and the wait times associated with those products. The operations research discipline of queueing theory has proven this mathematically (see the graphic) although few companies acknowledge this relationship as they seek greater operational efficiencies through higher utilization rates. <i>“Even though there’s ample capacity in this distribution center during most of the year, Zara opened a [new logistics center]… [because] Zara’s senior managers follow a fundamental rule of queuing models, which holds that waiting time shoots up exponentially when capacity is tight and demand is variable. By tolerating lower capacity utilization in its factories and distribution centers, Zara can react to peak or unexpected demands faster than its rivals.“</i> (Ferdows) They effectively sacrifice capacity to reduce the amount of inventory. (Schmidt)<br />
<br />
System Level Optimization: Zara also focuses on optimizing the system as opposed to local operations. <i>“Few managers can imagine sending a half-empty truck across Europe, paying for airfreight twice a week to ship coats on hangers to Japan, or running factories for only one shift.“</i> (Ferdows) Just as most businesses would try to maximize the amount of inventory in their distribution center, rather than minimize it.<br />
<br />
Inventory Allocation Algorithms: Zara partnered with faculty from MIT’s Operations Research department a few years ago to optimize the allocation of inventory across the store network. This involves regression forecasting the demand for each product at every store location, and then routing the items to the stores where they are the most likely to sell based on independent poisson process forecasts. The model <i>“moves excess inventory away from low‐selling stores where it is not needed, and send it instead to high‐performing stores where it thus reduces missed sales due to stock‐outs.”</i> (Caro) This also incorporates item requests from store managers, a ‘saturation effect’ to incorporate the amount of that product already sold in that location, actual sales, and auto-corrects for the credibility of store manager predictions. (Caro)<br />
<br />
Small Batch Size: <i>“Rather than chase economies of scale, Zara manufactures and distributes products in small batches.“</i> (Ferdows) This inspires customers to buy something if they like it, because it won’t be there later. (Ghemawat) This ‘climate of scarcity’ was deliberately created, on the assumption that customers can always find a different product that they would like to buy because of the greater variety of products and greater inventory turnover. (Ghemawat) (Caro) (Ferdows) This stands in sharp contrast to existing methods of batch size optimization such as Economic Order Quantity (Youngman) which balance the trade-off between inventory costs and setup time.<br />
<br />
Instead, Zara sought advantage in smaller batch sizes than the industry for the associated speed benefits. <i>“In actuality, reducing batch size simultaneously reduces queue time and wait time”</i>, <i>“[Reducing batch sizes doesn’t speed throughput] by speeding up machine or process time, but by reducing idle time when work sits on the workshop floor between process points.”</i> (Youngman) Smaller batch sizes also imply that setup costs have been greatly reduced, if not eliminated. Otherwise, producing the variety of goods (44,000 items per year) would be cost prohibitive. This can alternatively be viewed as causing, <i>“…operator skills atrophy (3), decreasing set-up frequency drives set-up proficiency down.”</i> (Youngman) Perhaps Zara’s large variety of production gave them a tactical advantage in setup configuration and speed, which translated into a unique business model.<br />
<br />
<b>Non-quantitative Methods</b><br />
<br />
Speed to Market: Zara can design and deliver a garment in 15 days, compared to several months at their competitors. (Ferdows) One of these speed improvements is highlighted in their decision to overnight air freight all deliveries despite the cost involved. Being vertically integrated also gives them a 40% speed advantage by avoiding the wholesale purchasing process. <i>“Trends that fail are ditched, while production of those that do work can be increased.”</i> (Hall)<br />
<br />
Speed to market also reduces their working capital requirements, because inventory is rarely in transit and because they turn over their inventory 4x more often than other apparel retailers. (Ghemawat) The greatest advantage though, is that Zara can see the designs on the runway, and get the apparel in their stores weeks before the designer can. (Ferdows)<br />
<br />
Manage Product Lines Independently to Avoid Bottlenecks: Zara operates the women’s, men’s, and children’s clothing lines almost completely independently (some people have multiple roles at smaller stores). <i>“Though it’s more expensive to operate three channels, the information flow for each channel is fast, direct, and unencumbered by problems in other channels…“</i> (Ferdows)<br />
<br />
<br />
<b>Quantification of Benefits</b><br />
<div class="separator" style="clear: both; text-align: center;">
<a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEj8I_z_lqPIHH93zC9rSTIjwkstzkXwyC7urhREXP33iWW3AYTzXa2aqfQ1OD2oruTzGh3-nGoJGCQxpSWjGl4ztWjw9WBMkAYhRdOhEV5H6N1Nbp_wVnzR14qp9IvfruiyX_gi9VSRA9rR/s1600/Zara+Growth.png" imageanchor="1" style="clear: left; float: left; margin-bottom: 1em; margin-right: 1em;"><img border="0" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEj8I_z_lqPIHH93zC9rSTIjwkstzkXwyC7urhREXP33iWW3AYTzXa2aqfQ1OD2oruTzGh3-nGoJGCQxpSWjGl4ztWjw9WBMkAYhRdOhEV5H6N1Nbp_wVnzR14qp9IvfruiyX_gi9VSRA9rR/s1600/Zara+Growth.png" /></a></div>
<br />
• Location Growth: 21% CAGR in locations over the past 14 years, <i>“while like-for-like sales growth had averaged 9% per year recently,“</i> (Ghemawat)<br />
• Revenue Growth: <i>“From 1991 to 2003, Inditex’s sales… grew more than 12-fold from €367 million to €4.6 billion...“</i> (Ferdows)<br />
• "<i>[Zara] overtook Gap [in 2008] as the number one fashion retailer.”</i> (Hall)<br />
• Implementing sophisticated distribution management algorithms <i>“increased sales by an estimated 3‐4%, corresponding to an [impact] of approximately $233M and $353M in additional revenues for 2007 and 2008, respectively.”</i> (Caro) This model also reduced shipments between stores, and it freed Zara from increasing the size of their Warehouse Allocation team as the business grew.<br />
• Discount Avoidance: Zara collects <i>“85% of the full ticket price on its retail clothing, while the industry average is 60% to 70%.”</i> (Ferdows) (Hall)<br />
• Speed to Market: <i>“Zara isn’t just a bit faster than rivals such as Gap, whose lead time is 9 months, it is 12 times faster.”</i> (IT)<br />
• Less Inventory: <i>“Zara has shown that… it can carry less inventory (about 10% of sales, compared to 14% to 15% at Benetton, H&M, and Gap).“</i> (Ferdows)<br />
• Unsold Inventory: <i>“In fact, Zara has an informal policy of moving unsold items after two or three weeks. This can be an expensive practice for a typical store, but since Zara stores receive small shipments and carry little inventory, the risks are small; unsold items account for less than 10% of stock, compared with the industry average of 17% to 20%."</i> (Ferdows) (Deschamps)<br />
• <a href="http://en.wikipedia.org/wiki/Bullwhip_effect">Bullwhip Effect</a>: <i>“The constant flow of updated data mitigates the so-called bullwhip effect—the tendency of supply chains… to amplify small disturbances… In an industry that traditionally allows retailers to change a maximum of 20% of their orders once the season has started, Zara lets them adjust 40% to 50%. In this way, Zara avoids costly overproduction and the subsequent sales and discounting prevalent in the industry.”</i> (Ferdows)<br />
• Failure Avoidance: Zara would wait to see customer demand for certain products before ramping up production. <i>“As a result, failure rates on new products only 1%, compared with an average of 10% for the sector.”</i> (Ghemawat)<br />
<br />
<br />
<b>Commentary</b><br />
<br />
Disclosure: I was dumbfounded that Zara permitted MIT to disclose the $353 million dollar impact of the inventory allocation project. This seems like a valuable trade secret, but I presume that in their first dealings with academics they failed to restrict what was disclosed in the publication of their results. Zara’s is also collaborating with MIT on optimization of their purchasing and pricing processes, although the financial impact of those efforts is not disclosed. In all likelihood, because Zara’s locked down the publication of such important information in their subsequent collaborations.<br />
<br />
Vertical Integration: Optimizing the supply chain in the apparel industry seems to require vertical integration. In particular, the wholesale purchasing process makes it difficult to rapidly change inventory estimates because larger batch sizes are needed to make the sales overhead a smaller component of each item. The bullwhip effect also leads each participant in the supply chain to overproduce so as to maximize local profits, while Zara deliberately leaves demand unsatisfied. This unsatisfied demand makes production more efficient, drastically reduces the expense of unloading unpopular items, and discourages planned subsequent purchasing by consumers.<br />
<br />
Advertising: Zara spends less than 1% on advertising vs. 4% for their competitors. This money goes straight to the bottom line, which may help in part to explain their profitability. (Hall) (Ghemawatt) This is a difference in advertising strategy that gives them larger margins, but I haven’t heard of any experiments or testing that determine whether the traditional apparel advertising strategy is optimal so we just need to recognize this difference.<br />
<br />
Logistics, Communications, Battles: I’ve quoted Patton before, <i>“Wars are won by logistics, communications, and battles, in that order.”</i> It is therefore interesting to see Zara dominate it’s industry based on logistics prowess, but to also emphasize communications in their operations. <i>“Consider this detail: The sales clerks in Zara are obliged to communicate the requests of their customers [to management], including even the suggestions that customers make in fitting rooms,”</i> (Wharton) They deliberately segregate the operations by the target consumers (women, men, children) so prevent a bottleneck in any one line from affecting others.<br />
<br />
Speed is Paramount: <i>“The Zara model may be unique, but at its heart is a perfectly simple principle: In fashion, nothing is as important as time to market – not advertising (which Zara does just twice a year in newspapers), not sales promotions (which Zara does only sparingly), not even labor costs.”</i> (IT) It is therefore open to questioning, how much of Zara’s success is due to the single minded pursuit of speed to market vs. the deliberate engineering of their supply chain.<br />
<br />
Always Simulate First: To quantify the impact of the inventory allocation models, they ran parallel testing to model the financial impacts. This represents a best practice, with both parallel testing and simulation comprising key steps in rollout best practices.<br />
<br />
<br />
<b>Works Cited</b><br />
<span style="font-size: x-small;"><br /></span>
<span style="font-size: x-small;"><a href="http://www.telegraph.co.uk/finance/newsbysector/retailandconsumer/2794912/Zara-is-now-bigger-than-Gap.html">“Zara is now bigger than the Gap.”</a> By James Hall. The Telegraph. August 17, 2008.</span><br />
<span style="font-size: x-small;">http://www.telegraph.co.uk/finance/newsbysector/retailandconsumer/2794912/Zara-is-now-bigger-than-Gap.html</span><br />
<span style="font-size: x-small;"><br /></span>
<span style="font-size: x-small;"><a href="http://web.mit.edu/jgallien/www/ZaraInterfacesPaperDraftFeb23.pdf">“Zara Uses Operations Research to Reengineer Its Global Distribution Process.</a>” By Felipe Caro, Jérémie Gallien, Javier García Torralbo, Jose Manuel Corredoira Corras, Marcos Montes Vazquez, José Antonio Ramos Calamonte, and Juan Correa. February 23, 2009. </span><br />
<span style="font-size: x-small;">http://web.mit.edu/jgallien/www/ZaraInterfacesPaperDraftFeb23.pdf</span><br />
<span style="font-size: x-small;"><br /></span>
<span style="font-size: x-small;">Wikipedia: <a href="http://en.wikipedia.org/wiki/Fast_fashion">Fast Fashion</a>. December 20th, 2013.</span><br />
<span style="font-size: x-small;">http://en.wikipedia.org/wiki/Fast_fashion</span><br />
<span style="font-size: x-small;"><br /></span>
<span style="font-size: x-small;"><a href="https://elearning.ec.unipi.it/claroline/backends/download.php?url=L1phcmEucGRm&cidReset=true&cidReq=PP151_002">“Zara: Fast Fashion.”</a> By Pankaj Ghemawat and Jose Luis Nueno. December 21, 2006. Harvard Business School Case.</span><br />
<span style="font-size: x-small;">https://elearning.ec.unipi.it/claroline/backends/download.php?url=L1phcmEucGRm&cidReset=true&cidReq=PP151_002</span><br />
<span style="font-size: x-small;"><br /></span>
<span style="font-size: x-small;"><a href="http://blogs.eafit.edu.co/gec/files/2010/07/Zara_HBR_Rapid_Fire_Fulfillment.pdf">“Rapid-Fire Fulfillment.”</a> By Kasra Ferdows, Michael A. Lewis, and Jose A.D. Machuca. Harvard Business Review, November 2004 Issue.</span><br />
<span style="font-size: x-small;">http://blogs.eafit.edu.co/gec/files/2010/07/Zara_HBR_Rapid_Fire_Fulfillment.pdf</span><br />
<span style="font-size: x-small;"><br /></span>
<span style="font-size: x-small;"><a href="http://www.cob.sjsu.edu/aggarwal_n/teaching/classes/Material188/Zara%20Case.pdf">“Zara and Inditex. Using Information Technology for Competitive Advantage.”</a></span><br />
<span style="font-size: x-small;">http://www.cob.sjsu.edu/aggarwal_n/teaching/classes/Material188/Zara%20Case.pdf</span><br />
<span style="font-size: x-small;"><br /></span>
<span style="font-size: x-small;"><a href="http://www.neilsonjournals.com/OMER/sOMTriangle.pdf">“The OM Triangle.”</a> By Glenn Schmidt. 2005/ Operations Management Education Review 1: 87-104. Neilsen Journals Publishing.</span><br />
<span style="font-size: x-small;">http://www.neilsonjournals.com/OMER/sOMTriangle.pdf</span><br />
<span style="font-size: x-small;"><br /></span>
<a href="http://www.economist.com/node/21551063" style="font-size: small;">“Fashion Forward: Zara, Spain’s most successful brand, is trying to go global.”</a><span style="font-size: x-small;"> The Economist. March 24th, 2012.</span><br />
<span style="font-size: x-small;">http://www.economist.com/node/21551063</span><br />
<span style="font-size: x-small;"><br /></span>
<span style="font-size: x-small;"><a href="http://dianeisabelle.com/2012/09/08/zaras-business-model-and-competitive-advantages/">“Zara’s Business Model and Competitive Advantages.”</a> By Diana Isabelle. September 8th, 2012. </span><br />
<span style="font-size: x-small;">http://dianeisabelle.com/2012/09/08/zaras-business-model-and-competitive-advantages/</span><br />
<span style="font-size: x-small;"><br /></span>
<span style="font-size: x-small;"><a href="http://knowledge.wharton.upenn.edu/article/inditex-dazzles-its-competitors-again/">“Inditex Dazzles Its Competitors Again.”</a> Knowledge At Wharton. November 2, 2005.</span><br />
<span style="font-size: x-small;">http://knowledge.wharton.upenn.edu/article/inditex-dazzles-its-competitors-again/</span><br />
<span style="font-size: x-small;"><br /></span>
<span style="font-size: x-small;"><a href="http://www.just-style.com/management-briefing/fast-fashions-competitive-advantages_id114806.aspx">“just-style management briefing: Fast Fashion’s competitive advantages.”</a> By MS Deschamps. July 2nd, 2012.</span><br />
<span style="font-size: x-small;">http://www.just-style.com/management-briefing/fast-fashions-competitive-advantages_id114806.aspx</span><br />
<span style="font-size: x-small;"><br /></span>
<span style="font-size: x-small;"><a href="http://dspace.mit.edu/bitstream/handle/1721.1/52779/526744773.pdf?sequence=1">“Clearance Pricing Optimization at Zara.”</a> By Rodolfo Carboni Borose. Published by Massachussettes Institute of Technology. 2009.</span><br />
<span style="font-size: x-small;">http://dspace.mit.edu/bitstream/handle/1721.1/52779/526744773.pdf?sequence=1</span><br />
<span style="font-size: x-small;"><br /></span>
<span style="font-size: x-small;"><a href="http://dspace.mit.edu/handle/1721.1/74454">“Demand Forecast for Short Life-cycle Products: Zara case study.”</a> By Tatiana Bonefoi. Published by Massachussettes Institute of Technology. 2010.</span><br />
<span style="font-size: x-small;">http://dspace.mit.edu/handle/1721.1/74454</span><br />
<span style="font-size: x-small;"><br /></span>
<span style="font-size: x-small;"><a href="http://www.dbrmfg.co.nz/Production%20Batch%20Issues.htm">“A Guide to Implementing the Theory of Constraints.”</a> Dr. K. J. Youngman</span><br />
<span style="font-size: x-small;">http://www.dbrmfg.co.nz/Production%20Batch%20Issues.htm</span><br />
<div>
<br /></div>
<div class="MsoNormal">
</div>
</div>
CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.comtag:blogger.com,1999:blog-4323599490342956532.post-85705658219546932252013-06-14T00:00:00.000-04:002013-06-14T00:03:21.155-04:00Porsche Lean Manufacturing and Statistical Process Control<div dir="ltr" style="text-align: left;" trbidi="on">
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<b>Background<o:p></o:p></b></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<span style="background-color: white; background-position: initial initial; background-repeat: initial initial;">Porsche in the 1990’s <i>“was fighting for its own survival.”</i> (<span class="SpellE">Plumer</span>) After orders decreased 70% from 1986 to 1993 </span><i>“the company was teetering on the verge of bankruptcy, and there were whispers about a possible takeover.”</i> (Henderson) <i>“Recession had crippled sales, and costs were out of control.”</i> (Nash) From such descriptions, <span class="SpellE">its</span> amazing that Porsche is now valued at $27 billion and makes more per car than any other auto manufacturer. <span style="background-color: white; background-position: initial initial; background-repeat: initial initial;"><o:p></o:p></span></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<span style="background-color: white; background-position: initial initial; background-repeat: initial initial;">The road to recovery for Porsche was based on <a href="http://en.wikipedia.org/wiki/Lean_manufacturing">lean manufacturing</a>, which the newly appointed CEO Dr. <span class="SpellE"><a href="http://en.wikipedia.org/wiki/Wendelin_Wiedeking">Wiedeking</a></span> was familiar with from his prior company. He brought in Japanese lean manufacturing consultants as quickly as possible (nearly all of them from Toyota) and they began converting Porsche into a <a href="http://en.wikipedia.org/wiki/Just_in_time_(business)">JIT</a> enterprise. Managerial staff were reduced by 30%, reassigned, and willfully submitted to reprimands from the consultants. They succeeded in reducing Porsche’s costs, improving throughput, and eliminating excess inventory but this only tells part of the story.<o:p></o:p></span></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<span style="background-color: white; background-position: initial initial; background-repeat: initial initial;">Part of the story is that lean manufacturing was built on principles of quality manufacturing developed by W. Edwards Deming “The Man Who Taught the Japanese Quality”. <span class="GramE">One of his central tenets being statistical process control, which I regard as the quantitative method at the core of lean manufacturing and Porsche’s turnaround.</span><o:p></o:p></span></div>
<div class="MsoNormal" style="background-color: white; font-family: Calibri, sans-serif; font-size: 11pt; line-height: 16.5pt; margin: 0in 0in 7.5pt 0.5in;">
<span style="font-size: 9pt;">"The traditional craftsmanship for which Germany became famous was filing and fitting parts so that they fit perfectly," Professor Jones said. "But that was wasted time. The parts should have been made right the first time. So the new craftsmanship is the craftsmanship of thinking up clever ways of making things simpler and easier to assemble. It is the craft of creating an uninterrupted flow of manufacturing." (Nash)<o:p></o:p></span></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
The phrase <i>“filing and fitting parts so that they fit perfectly”</i> clearly indicates that the production line was not statistically in control, otherwise parts would not need to be filed and fitted. <span class="GramE">Time to complete specific job line tasks would also vary wildly “as workers would sometimes climb ladders to look for parts” again pointing to a process not in statistical control.</span></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
The tale of the Porsche turnaround is therefore a story of lean manufacturing AND Porsche’s introduction to Deming’s quality management principles. For example, <i>“You can’t inspect quality into a product.”</i> It is there or is absent by the time it is inspected.</div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<br /></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<b>Non-Quantitative Methods: Lean Manufacturing<o:p></o:p></b></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<span style="background-color: white; background-position: initial initial; background-repeat: initial initial;">The lean manufacturing methods that Porsche implemented get most of the credit in the press for the company’s transformation, although few of these are quantitative methods and all of them are just logical. </span><i>“On the first day the consultants arrived, they cut all of the shelves in half.”</i> (Nash) The consultants also said that the factory looked more like a warehouse, with the large number of shelves for inventory. The significance of halving inventory may be lost on non-industrial engineers, so I should point out that inventory requires money to buy and is expensive to have large amounts of it. It also exposes companies to obsolescence risks if there is a drop in demand for a product. Viewed equivalently, <span class="GramE">all of the</span> inventory sitting on shelves was exactly the same as having stacks of hundred dollar bills on all of the shelves for a month rather than sitting in a bank account. Warehousing is also expensive because space costs money. You need to lease the land, heat the storage space, buy a roof to keep the weather out, pay security to ensure no one breaks in, etc. It is also less chaotic to have only the parts that you need on the factory floor so that you’re not tripping over unneeded parts when trying to complete a task. If you maintain minimal inventory these expenses all vanish.</div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<span style="background-color: white; background-position: initial initial; background-repeat: initial initial;">Now some of you are assuredly asking, but what if you don’t have a part when you need it? Part of the answer is creating a JIT or ‘Just <span class="GramE">In</span> Time’ delivery system, whereby suppliers agree to deliver parts no earlier than 24 hours before they will be used. The next part of this process is ensuring that parts do not arrive to a station on the production line before they are needed. This is handled with a <span class="SpellE"><a href="http://en.wikipedia.org/wiki/Kanban">kanban</a></span> ‘card’ system which matches inventory requests to the specific job and position on the production line of that job. <span class="GramE">By controlling the number of ‘cards’ on the production line and forbidding the release of inventory without a card, the amount of inventory is minimized.</span><o:p></o:p></span></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<span style="background-color: white; background-position: initial initial; background-repeat: initial initial;">Now some readers may be asking, “But how does JIT work if bad weather prevents the delivery of some parts?” This is a valid question that many people have asked. The counterintuitive truth though, is that the <span class="GramE">costs of warehousing excess inventory are</span> a hundred of times more expensive than a delay caused by a missing part. <o:p></o:p></span></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<br /></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<b><span style="background-color: white; background-position: initial initial; background-repeat: initial initial;">Quantitative Methods: Deming and Statistical Process Control<o:p></o:p></span></b></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<span style="background-color: white; background-position: initial initial; background-repeat: initial initial;">The first matter that the new CEO pursued <i>“was to benchmark every aspect of production to find out how much time, effort, and money <span class="GramE">was</span> being spent on making a Porsche.”</i> (<span class="SpellE">Mudd</span>) This is a classic Deming principle, that <i>“You cannot manage what you cannot measure”</i> and is easily recognized in the plotting of confidence intervals, trend charts, and segregation of processes as ‘in control’ <span class="SpellE">vs</span> ‘out of control’.<o:p></o:p></span></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<span class="SpellE"><span style="background-color: white; background-position: initial initial; background-repeat: initial initial;">Wiedeking</span></span><span style="background-color: white; background-position: initial initial; background-repeat: initial initial;"> also created the <i>“Porsche Improvement Program, a program designed to measure quality and efficiency and eliminate waste”</i> (<span class="SpellE">Mudd</span>) and its description could well be used as a description of Deming’s lifelong accomplishments or the results of his methods. It’s also noteworthy that consultants were brought in from Japan, probably the one country which adopted the tenets of Deming’s philosophy most faithfully.<o:p></o:p></span></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<span style="background-color: white; background-position: initial initial; background-repeat: initial initial;">Rounding out the overlaps between Porsche and Deming, are that consultants were sent out to Porsche’s suppliers to train them in the techniques (an idea first popularized by Deming) (Nash) and an organizational focus on trust. <i>“<span class="SpellE">Wiedeking</span> views the experience of remaking Porsche as a triumph of trust within an organization. It's a theme he comes back to again and again in the course of an interview. ‘With your workforce, with your employees, as well as with your <span class="GramE">shareholders, . . .</span> you must build some kind of trust,’ he says.”</i> (<span class="SpellE">Mudd</span>)<o:p></o:p></span></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<br /></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<b>Quantitative Methods: Pooling Principle and Simulation<o:p></o:p></b></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
A brief mention of the pooling principle is also in order. Porsche (which owns Volkswagen) deliberately designed the Porsche Cayenne and Volkswagen <span class="SpellE">Touareg</span> to share the same chassis. (<span class="SpellE">Mudd</span>) Explicit calculations in <span class="SpellE">queueing</span> theory prove that this is more efficient than establishing two production lines because it permits random lulls in demand for one model to be naturally offset by random peaks in demand for another. The pooling of production lines helps to deal with demand uncertainty.</div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<span style="background-color: white; background-position: initial initial; background-repeat: initial initial;">Engineers also shifted much of their design to computer simulations as a way of cutting the time for prototyping in half. (<span class="SpellE">Mudd</span>) This has follow-on benefits in that it eliminates raw material costs for prototyping, rapid iteration, and translates well into CAD based manufacturing trends.<o:p></o:p></span></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<br /></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<b>Quantification of Benefits<o:p></o:p></b></div>
<div class="MsoListParagraphCxSpFirst" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 0.0001pt 0.5in; text-indent: -0.25in;">
<span style="font-family: Symbol;">·<span style="font-family: 'Times New Roman'; font-size: 7pt; line-height: normal;"> </span></span>On arrival to the company, <span class="SpellE">Wiedeking</span> promised the board that he could reduce expenses by 30%, and delivered.</div>
<div class="MsoListParagraphCxSpMiddle" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 0.0001pt 0.5in; text-indent: -0.25in;">
<span style="font-family: Symbol;">·<span style="font-family: 'Times New Roman'; font-size: 7pt; line-height: normal;"> </span></span><i>“By 2007, Porsche was the world’s most profitable automaker on a per unit basis.”</i> (Henderson)</div>
<div class="MsoListParagraphCxSpMiddle" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 0.0001pt 0.5in; text-indent: -0.25in;">
<span style="font-family: Symbol;">·<span style="font-family: 'Times New Roman'; font-size: 7pt; line-height: normal;"> </span></span><i>“Porsche, despite also suffering a sales slump, report a nearly $11 billion profit in the last half of 2008.”</i> (<span class="SpellE">Plumer</span>)<span style="background-color: white; background-position: initial initial; background-repeat: initial initial;"><o:p></o:p></span></div>
<div class="MsoListParagraphCxSpMiddle" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 0.0001pt 0.5in; text-indent: -0.25in;">
<span style="font-family: Symbol;">·<span style="font-family: 'Times New Roman'; font-size: 7pt; line-height: normal;"> </span></span><i>“As <span class="SpellE">Wiedeking</span> explained, ‘<span class="GramE">Our</span> sales doubled in just six years as did its revenue through organic growth.’”</i> (Henderson)</div>
<div class="MsoListParagraphCxSpMiddle" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 0.0001pt 0.5in; text-indent: -0.25in;">
<span style="font-family: Symbol;">·<span style="font-family: 'Times New Roman'; font-size: 7pt; line-height: normal;"> </span></span>Improvements in operational metrics yield higher profit per car. (Henderson)</div>
<div class="MsoListParagraphCxSpMiddle" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 0.0001pt 1in; text-indent: -0.25in;">
<span style="font-family: 'Courier New';">o<span style="font-family: 'Times New Roman'; font-size: 7pt; line-height: normal;"> </span></span>Inventory reduced from 28 days to 1 <span class="GramE">days</span>. (Nash) (<span class="SpellE">Mudd</span>)</div>
<div class="MsoListParagraphCxSpMiddle" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 0.0001pt 1in; text-indent: -0.25in;">
<span style="font-family: 'Courier New';">o<span style="font-family: 'Times New Roman'; font-size: 7pt; line-height: normal;"> </span></span>Assembly time reduced from 120 hours to 60 hours. (Nash) (<span class="SpellE">Mudd</span>)</div>
<div class="MsoListParagraphCxSpMiddle" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 0.0001pt 1in; text-indent: -0.25in;">
<span style="font-family: 'Courier New';">o<span style="font-family: 'Times New Roman'; font-size: 7pt; line-height: normal;"> </span></span><i>“Errors per car have fallen 50 percent”</i> (Nash)</div>
<div class="MsoListParagraphCxSpMiddle" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 0.0001pt 1in; text-indent: -0.25in;">
<span style="font-family: 'Courier New';">o<span style="font-family: 'Times New Roman'; font-size: 7pt; line-height: normal;"> </span></span>Work forced reduced 19 percent. (Nash)</div>
<div class="MsoListParagraphCxSpMiddle" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 0.0001pt 1in; text-indent: -0.25in;">
<span style="font-family: 'Courier New';">o<span style="font-family: 'Times New Roman'; font-size: 7pt; line-height: normal;"> </span></span><i>“Factory space has been reduced by 30 percent.”</i> (Nash)</div>
<div class="MsoListParagraphCxSpMiddle" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 0.0001pt 1in; text-indent: -0.25in;">
<span style="font-family: 'Courier New';">o<span style="font-family: 'Times New Roman'; font-size: 7pt; line-height: normal;"> </span></span>‘Concept to Launch’ reduced from 7 years to 3 years. (<span class="SpellE">Plumer</span>)</div>
<div class="MsoListParagraphCxSpMiddle" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 0.0001pt 1in; text-indent: -0.25in;">
<span style="font-family: 'Courier New';">o<span style="font-family: 'Times New Roman'; font-size: 7pt; line-height: normal;"> </span></span>Reduced throughput from 6 weeks per car to 3 days per car. (<span class="SpellE">Plumer</span>)</div>
<div class="MsoListParagraphCxSpMiddle" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 0.0001pt 1in; text-indent: -0.25in;">
<span style="font-family: 'Courier New';">o<span style="font-family: 'Times New Roman'; font-size: 7pt; line-height: normal;"> </span></span>Parts originally delivered 3 days late are now delivered JIT. (<span class="SpellE">Plumer</span>)</div>
<div class="MsoListParagraphCxSpLast" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt 0.5in; text-indent: -0.25in;">
<span style="font-family: Symbol;">·<span style="font-family: 'Times New Roman'; font-size: 7pt; line-height: normal;"> </span></span><i>“</i><i>Porsche survived [near bankruptcy] -- not only as a cachet name in German automobiles, but as the last remaining independent European manufacturer of sports cars.”</i> (Nash)</div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<br /></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<b>Personal Notes<o:p></o:p></b></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
I’m pleased to come across this example of quant CEO excellence. I’ve written about a correlation between company <span class="SpellE">overperformance</span> and quant CEOs <a href="http://cavqm.blogspot.com/2012/01/do-successful-analytics-companies.html">before</a>, and Porsche under Dr. <span class="SpellE">Wiedeking</span> (who earned a Doctorate in Engineering) is no exception.</div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
I also wanted to propound a few of Deming’s more theoretical points without substantiating them further.</div>
<div class="MsoListParagraphCxSpFirst" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 0.0001pt 0.5in; text-indent: -0.25in;">
<span style="font-family: Symbol;">·<span style="font-family: 'Times New Roman'; font-size: 7pt; line-height: normal;"> </span></span>There are two sources of variation, systemic and idiosyncratic, of which management <span class="GramE">can</span> only controls systemic variation.</div>
<div class="MsoListParagraphCxSpMiddle" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 0.0001pt 0.5in; text-indent: -0.25in;">
<span style="font-family: Symbol;">·<span style="font-family: 'Times New Roman'; font-size: 7pt; line-height: normal;"> </span></span>Root cause analysis should be pursued, so that the ‘Disease is cured, rather than the symptom.’</div>
<div class="MsoListParagraphCxSpLast" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt 0.5in; text-indent: -0.25in;">
<span style="font-family: Symbol;">·<span style="font-family: 'Times New Roman'; font-size: 7pt; line-height: normal;"> </span></span><i>“Price has no meaning without a measure of the quality being purchased.”<o:p></o:p></i></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<br /></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<b><span style="font-size: 9pt; line-height: 13px;">Citations<o:p></o:p></span></b></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<span style="font-size: 9pt; line-height: 13px;">“What’s Driving Porsche?” By Rebecca Henderson and Cate <span class="SpellE">Reavis</span><o:p></o:p></span></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<a href="https://mitsloan.mit.edu/LearningEdge/CaseDocs/08-075-What's%20Driving%20Porsche.Henderson.pdf" style="color: purple;"><span style="font-size: 9pt; line-height: 13px;">https://mitsloan.mit.edu/LearningEdge/CaseDocs/08-075-What's%20Driving%20Porsche.Henderson.pdf</span></a><span style="font-size: 9pt; line-height: 13px;"><o:p></o:p></span></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<br /></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<span class="GramE"><span style="font-size: 9pt; line-height: 13px;">“Back In High Gear.”</span></span><span style="font-size: 9pt; line-height: 13px;"> <span class="GramE">By Tom <span class="SpellE">Mudd</span>.</span> <span class="SpellE"><span class="GramE">IndustryWeek</span></span><span class="GramE">.</span> December 21, 2004.<o:p></o:p></span></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<a href="http://www.industryweek.com/companies-amp-executives/back-high-gear" style="color: purple;"><span style="font-size: 9pt; line-height: 13px;">http://www.industryweek.com/companies-amp-executives/back-high-gear</span></a><span style="font-size: 9pt; line-height: 13px;"><o:p></o:p></span></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<br /></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<span class="GramE"><span style="font-size: 9pt; line-height: 13px;">Porsche Consulting Company History.</span></span><span style="font-size: 9pt; line-height: 13px;"><o:p></o:p></span></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<a href="http://www.porscheconsulting.com/pco/en/company/history/" style="color: purple;"><span style="font-size: 9pt; line-height: 13px;">http://www.porscheconsulting.com/pco/en/company/history/</span></a><span style="font-size: 9pt; line-height: 13px;"><o:p></o:p></span></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<span style="font-size: 9pt; line-height: 13px;"> </span><span style="font-size: 9pt; line-height: 13px;"> </span></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<span style="font-size: 9pt; line-height: 13px;">“How Porsche has become one of Europe’s Most Successful Automakers.” <span class="GramE">By Robert Plume.</span> April 5, 2009.<o:p></o:p></span></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<a href="http://www.helium.com/items/1403075-how-porsche-has-become-one-of-europes-most-successful-automakers" style="color: purple;"><span style="font-size: 9pt; line-height: 13px;">http://www.helium.com/items/1403075-how-porsche-has-become-one-of-europes-most-successful-automakers</span></a><span style="font-size: 9pt; line-height: 13px;"><o:p></o:p></span></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<br /></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<span class="GramE"><span style="font-size: 9pt; line-height: 13px;">“Putting Porsche Back in the Pink.”</span></span><span style="font-size: 9pt; line-height: 13px;"> <span class="GramE">By Nathanial Nash.</span> January 20<sup>th</sup>, 1996. <span class="GramE">New York Times.</span><o:p></o:p></span></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<a href="http://www.nytimes.com/1996/01/20/business/putting-porsche-in-the-pink.html?pagewanted=all&src=pm" style="color: purple;"><span style="font-size: 9pt; line-height: 13px;">http://www.nytimes.com/1996/01/20/business/putting-porsche-in-the-pink.html?pagewanted=all&src=pm</span></a><span class="MsoHyperlink" style="color: blue; text-decoration: underline;"><span style="font-size: 9pt; line-height: 13px;"><o:p></o:p></span></span></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<br /></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<span class="GramE"><u><span style="font-size: 9pt; line-height: 13px;">Out of the Crisis.</span></u></span><span style="font-size: 9pt; line-height: 13px;"> <span class="GramE">By W. Edwards Deming.</span> <span class="GramE">Published 1982.</span> <span class="GramE">MIT Press.</span><o:p></o:p></span></div>
<div class="MsoNormal" style="font-family: Calibri, sans-serif; font-size: 11pt; line-height: 17px; margin: 0in 0in 10pt;">
<a href="http://www.amazon.com/Out-Crisis-W-Edwards-Deming/dp/0262541157" style="color: purple;">http://www.amazon.com/Out-Crisis-W-Edwards-Deming/dp/0262541157</a></div>
</div>
CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.comtag:blogger.com,1999:blog-4323599490342956532.post-5760708955758034902012-12-16T09:49:00.001-05:002012-12-16T09:49:08.615-05:00Narwhal, Dreamcatcher, Houdini, Optimizer, Call Tool and Blaster - The ‘MoneyBall Election’ of 2012<div dir="ltr" style="text-align: left;" trbidi="on">
<div class="MsoNormal">
The 2012 Presidential election has been labeled the
‘<a href="http://www.newyorker.com/online/blogs/newsdesk/2012/11/our-money-ball-election.html">Moneyball Election</a>’for the transformative role that big data and analytics played in the Obama
campaign’s victory. On the other hand though, <a href="http://twitter.com/fivethirtyeight">Nate Silver</a> of the <a href="http://fivethirtyeight.blogs.nytimes.com/">New York Times</a> accurately predicted how every state would vote in the election,
and accurately predicted 49 of 50 states in the 2008 election months before
election day.</div>
<div class="separator" style="clear: both; text-align: center;">
<a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhcRSa0683mTuuRfahG7ms41DbM4MhS8Im5AqT3Cuqim-UGl8jWmVcCuOCfV3jBya7yGTUqm9kaignhLVICTBorYqgUT4ZfRjfhU_cbV4UDJLdTRa7xs920nvbRLUp1pw3APJGn8O06VMw5/s1600/538+Obama+Romney+Winning+Likelihood.JPG" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><span 0="0" height="198" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhcRSa0683mTuuRfahG7ms41DbM4MhS8Im5AqT3Cuqim-UGl8jWmVcCuOCfV3jBya7yGTUqm9kaignhLVICTBorYqgUT4ZfRjfhU_cbV4UDJLdTRa7xs920nvbRLUp1pw3APJGn8O06VMw5/s320/538+Obama+Romney+Winning+Likelihood.JPG" style="color: black;" width="320"></span></a></div>
<div align="left" style="background: white; margin-bottom: 9.4pt; margin-left: 0in; margin-right: 0in; margin-top: 9.4pt; text-align: left;">
All his analysis did was to weight accurate polls more
heavily than they are in the media, and ignore controversial but historically
inaccurate polls. By doing so, he’s simultaneously proven how little the
political machine influences voter behavior.</div>
<div style="background: white; margin-bottom: 9.4pt; margin-left: 0in; margin-right: 0in; margin-top: 9.4pt;">
<i>“But, perhaps the most
devastating impact on traditional punditry: politics and campaigning has a
relatively small impact on elections. According to Silver’s model, Obama had a
strong likelihood of winning several months before the election. The economy picked
up before the election. Any conservative challenger had an uphill battle.” </i>(Ferenstein)</div>
<div class="MsoNormal">
So the question then becomes, how significant were the
technological additions to the political arsenal, and did any of them create a
competitive advantage?</div>
<br />
<div class="MsoNormal">
<b>Quantitative Methods</b></div>
<div class="MsoNormal">
Many technologies were deployed by the Obama campaign, but
only a portion of them were quantitative methods.
<br />
<ul>
<li><a href="http://cavqm.blogspot.com/2011/10/ab-testing-and-experimentation-as.html">A/B Testing</a>: Dozens
of tests performed per day to deliver incremental improvements in messaging,
fundraising and site engagement.</li>
<div class="separator" style="clear: both; text-align: center;">
<br /></div>
<ul>
<li>
<i>“</i><i>For the button, an <a href="http://www.wired.com/business/2012/04/ff_abtesting/all/">A/B test</a> </i><i>of three new word choices—”Learn More,” “Join Us
Now,” and “Sign Up Now”—revealed that “Learn More” garnered 18.6 percent more
signups per visitor than the default of “Sign Up.” Similarly, a black-and-white
photo of the Obama family outperformed the default turquoise image by 13.1
percent.” </i>(Wired)</li>
<li>
<i>“Almost
unanimously, staffers expected that a video of Obama speaking at a rally would
handily outperform any still photo. But in fact the video fared 30.3 percent
worse than even the turquoise image.”</i> (Wired)</li>
<li>
<i>“Many of
the e-mails sent to supporters were just tests, with different subject lines,
senders and messages.”</i> (Time)</li>
<li>
<i>“Any time you received an email from the Obama
campaign, it had been tested on 18 smaller groups and the response rates had
been gauged. The campaign thought all the letters had a good chance of
succeeding, but the worst-performing letters did only 15 to 20 percent of what
the best-performing emails could deliver. So, if a good performer could do $2.5
million, a poor performer might only net $500,000. The genius of the campaign
was that it learned to stop sending poor performers.”</i> (Atlantic)</li>
</ul>
<li>
<a href="http://www.slate.com/articles/news_and_politics/victory_lab/2012/01/project_dreamcatcher_how_cutting_edge_text_analytics_can_help_the_obama_campaign_determine_voters_hopes_and_fears_.html">Dreamcatcher</a>:
Text Mining and “Microlistening” (Slate)</li>
<ul>
<li>Call lists prioritized by persuadability, and bythe message channel</li>
<li>
<i>“What they revealed as they pulled back the
curtain was a massive data effort that helped Obama raise $1 billion, remade
the process of targeting TV ads and created detailed models of swing-state
voters that could be used to increase the effectiveness of everything from
phone calls and door knocks to direct mailings and social media.” </i>(Time)</li>
<li>
Sounds like they were relying on audience
descriptions similar to those produced by Blue Fin Labs to select the best
advertising slots. <i>“As a result, the
campaign bought ads to air during unconventional programming, like<span class="apple-converted-space"></span>'Sons of Anarchy',<span class="apple-converted-space"></span>'The Walking Dead'<span class="apple-converted-space"></span>and<span class="apple-converted-space"></span>'Don’t Trust the B—- in Apt. 23',
skirting the traditional route of buying ads next to local news programming.”</i>
(Time) </li>
<li><i>“Campaigns
do, however, take in plenty of information about what voters believe,
information that is not gathered in the form of a poll... As part of the
Dreamcatcher project, Obama campaign officials have already set out to redesign
the ‘notes’ field on individual records in the database they use to track
voters so that it sits visibly at the top of the screen—encouraging volunteers
to gather and enter that information.”</i> (Slate)</li>
</ul>
<li>Facebook Blaster and Twitter Blaster: Leveraged microtargeting
and volunteer’s the Facebook connections to promote voter registration of
likely Democrats. </li>
<ul>
<li>
<i>“The
digital, tech and analytics teams worked to build Twitter and Facebook Blasters…
With Twitter, one of the company's former employees, Mark Trammell, helped
build a tool that could specifically send individual users direct messages. ‘We
built an influence score for the people following the [Obama for America]
accounts and then cross-referenced those for specific things we were trying to
target, battleground states, that sort of stuff.’ Meanwhile, the teams also
built an opt-in Facebook outreach program that sent people messages saying,
essentially, ‘Your friend, Dave in Ohio, hasn't voted yet. Go tell him to
vote.’”</i> (Atlantic)</li>
</ul>
<li>
Optimizer: Let media buyers substitute cheaper
audiences for more expensive ones while advertising to the same target
audiences.</li>
<ul>
<li>
<i>“…allowed
the campaign to buy eyeballs on television more cheaply. They took set-top box
(that is to say, your cable or satellite box or DVR) data from Davidsen's old
startup, Navik Networks, and correlated it with the campaign's own data… Having
that data allowed the campaign to buy ads that they knew would get in front of
the most of their people at the least cost. They didn't have to buy the
traditional stuff like the local news, either. Instead, they could run ads
targeted to specific types of voters during reruns or off-peak hours.“</i>
(Atlantic)</li>
</ul>
<li>
<a href="http://www.blogger.com/[http://www.dailykos.com/story/2008/11/02/650137/-Obama-uses-Houdini]">Houdini</a>: A
project to let voters self-identify as democrats, and ensures that they get to
the polls in battleground states. </li>
<ul>
<li>
<i>“Then there was the much-vaunted secret
weapon, Project Houdini—a get-out-the-vote system that was supposed to
revolutionize the Election Day ground game. Each voter in each swing-state
voting precinct was assigned a numeric code; when poll watchers recorded the
voters arriving, the watchers were supposed to dial in the code to Houdini's
automated hotline.”</i>(Gallagher)</li>
</ul>
</ul>
</div>
<br />
<div class="MsoNormal">
<b>Non-Quantitative Methods (But Foundational to CAvQM)</b>
<br />
<li><a href="http://en.wikipedia.org/wiki/Project_Narwhal">Narwhal</a>:
An API and cloud-based data warehouse that facilitated real-time synchronous
information sharing between the campaign apps, field offices, and volunteer
efforts. Eliminated the traditional ‘siloed’ approach to field office
management and app development, and let all offices share the same
geographic/demographic profiling data. Simplified the development of apps and
analytics while reducing cost.</li>
<ul>
<li><i>“Allowed individual app scaling”</i> (Gallagher)</li>
<li><i>“Share one common data store”</i> (Gallagher)</li>
<li>
Call lists reconciled with fundraising lists,
field workers, social media and marketing databases. (Time)
</li>
</ul>
<a href="http://www.youtube.com/watch?v=lqSGFnOwQkM">Dashboard</a>:
Let’s anyone log in to the Obama campaign and immediately be given tasks to accomplish
based on real-time data. <i>“A ‘virtual
field office’”</i> (Gallagher)</div>
<ul>
<li><i>“Helped automate recruitment and outreach of volunteers.”</i> (Gallagher)</li>
</ul>
<div class="separator" style="clear: both; text-align: center;">
<a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhXJKDibsJmywhR9AAeKciWWnlEkn7p0UHC8I1ncJhtfE7zeS_vECdioBcAP5LnywtxDfkUrFNozFUtqOtYndXCYvuWPDXBaY7fKWFyeoTPftqH0pGY6upU_rF5Sp9vIDsOJ5JMliDj3vP4/s1600/Obama+Dashboard.png" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" height="201" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhXJKDibsJmywhR9AAeKciWWnlEkn7p0UHC8I1ncJhtfE7zeS_vECdioBcAP5LnywtxDfkUrFNozFUtqOtYndXCYvuWPDXBaY7fKWFyeoTPftqH0pGY6upU_rF5Sp9vIDsOJ5JMliDj3vP4/s400/Obama+Dashboard.png" width="400" /></a></div>
<br />
<br />
<li><a href="http://www.youtube.com/watch?v=lqSGFnOwQkM">Call Tool</a>:
<i>“It
allowed volunteers anywhere to join a call campaign, presenting a random
person's phone number and a script with prompts to follow. Call Tool also
allowed for users to enter notes about calls that could be processed by ‘collaborative
filtering’ on the back end—identifying if a number was bad, or if the person at
that number spoke only Spanish, for instance—to ensure that future calls were
handled properly.”</i> (Gallagher) </li>
<li><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgxHm-gd24UPkyGCd8XLwBepx23sNOCwZJccX6bgZbpO7DyJVzPY8E6w0Ou5A8k9_YzjXh1z3llsO5hRpJwJWbu8arQLGjhLuaK6pJyT8T9YfyxQfuWFrjZdwjwvId8LJ7xuQC1yduGSkZ0/s1600/Obama+Call+Tool.jpg" imageanchor="1" style="clear: right; float: center; margin-bottom: 1em; margin-left: 1em; text-align: center;"><img border="0" height="228" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgxHm-gd24UPkyGCd8XLwBepx23sNOCwZJccX6bgZbpO7DyJVzPY8E6w0Ou5A8k9_YzjXh1z3llsO5hRpJwJWbu8arQLGjhLuaK6pJyT8T9YfyxQfuWFrjZdwjwvId8LJ7xuQC1yduGSkZ0/s320/Obama+Call+Tool.jpg" width="320" /></a></li>
<li>Identity: This program gamified all digital volunteer
efforts and let them compete against one another.</li>
<ul>
<li><i>“Both
Call Tool and Dashboard—as well as nearly all of the other volunteer-facing
applications coded by the Obama campaign's IT team—integrated with another
application called Identity. Identity was a single-sign-on application that
tracked volunteer activity across various activities and allowed for all sorts
of campaign metrics, such as tracking the number of calls made with Call Tool
and displaying them in Dashboard as part of group "leaderboards." The
leaderboards were developed to "gamify" activities like calling,
allowing for what Ecker called "friendly competition" within groups
or regions.”</i> (Gallagher)</li>
</ul>
<br />
<br />
<div>
<b>Quantification of Benefits</b>
<br />
<ul>
<li>
Lower Cost Per Impression:<i> “Obama's campaign's cost per ad was lower ($594) than the Romney
campaign ($666) or any other major buyer in the campaign cycle.”</i> (The
Atlantic) That’s particularly significant when campaigns spend more than $500
million on television ads.</li>
<li>Fundraising Optimization:<i> “By the end of the [1998] campaign, it was estimated that a full 4
million of the 13 million addresses in the campaign’s email list, and some $75
million in money raised, resulted from [A/B testing].”</i> (Christian)</li>
<li>Lift in Response Rate:<i> “The campaign thought all the letters had a good chance of succeeding,
but the worst-performing letters did only 15 to 20 percent of what the
best-performing emails could deliver.” </i>(Atlantic)<i></i>
</li>
<li>Efficacy of Social Media: <i>“The campaign found that roughly 1 in 5 people contacted by a Facebook
pal acted on the request, in large part because the message came from someone
they knew.” </i>(Time)<i></i></li>
</ul>
</div>
<br />
<br />
<li>The Obama campaign ran <i>“the most tweeted tweet”</i> (The Atlantic)
</li>
The Obama campaign ran <i>“the most popular Facebook post”</i> (The Atlantic)
<li>Obama won. Twice. </li>
<br />
<br />
<div>
<b>Personal Notes</b></div>
<div>
<i>Political Analytics as Table Stakes:</i> Obama’s analytics department in 2012 was
five times the size of the 2008 operation. (Time) They created many powerful
additions to the political arsenal, but none influenced the outcome of the
election. To paraphrase Silver, incumbents always have an advantage, the
competitor wasn’t regarded highly even within his own party and the economy
began to improve months before the election. It’s almost a shame that both
political parties are sure to invest in such technologies before the next
elections, thereby degrading the competitive advantage.<br />
<br />
<i>Incumbency Advantage
Increases:</i> Incumbency advantage now translates into IT preparedness and the
greater technological sophistication of their campaigns. “<i>‘I don't think we would have
been able to [build Narwhal] if we had to deal with the primaries,’ Reed said.”</i> (Gallagher) Also, the Obama campaign asked
donors in 2008 for permission to use their information in future campaigns for
payment automation. I imagine we’ll see parties taking control for the
centralization of data and IT within two years, and it does beg the question,
Why didn’t anyone think of this before?<br />
<br />
<i>Narwhal as Real-time Communications
Advantage:</i> I didn’t get
Narwhal at first. In fact, it was weeks before I understood the significance of
Narwhal and the election was already over. Narwhal was useful for efficiency,
but it is powerful because it <u>replaces communication</u>. No communication
was wasted describing what had been done, who had been called already, etc.
This communication also <u>enabled dynamic resource allocation</u>. Prior
campaigns only had daily updates, meanwhile Narwhal updated in real time.</div>
<br />
<div align="center" class="MsoNormal" style="text-align: center;">
<span><u><b>Look for opportunities to automate
communication with APIs.</b></u></span></div>
<br />
<div>
<i>Centralization of IT Development:</i> If your IT systems are
fragmented with massive duplication of effort, IT centralization will be a
substantial improvement. <i>“’</i><i>
One of the biggest problems in the last
campaign was that you had all these people who are out in the field who are
volunteering who start building their own versions of these rogue tools to do
the same thing over and over again,’ said Clint Ecker, senior engineer for
Obama for America. Every field office assembled its own patchwork of tools
using spreadsheets or a hacked Web application to track operations. They
communicated over Google groups or simple e-mail lists. ‘It made it hard to
keep everyone on the same page,’ he added.”</i>
(Gallagher)</div>
<br />
<i>Data Driven Decision Making:</i> A <a href="http://cavqm.blogspot.com/2012/05/data-driven-decision-making-dddm-as.html">prior blog</a> defined DDDM as a competitive
advantage and the Obama campaign leveraged it fully. <i>“A</i><i>ssumptions were rarely left in place without
numbers to back them up.”</i> (Time) as evidenced by the heavy use of A/B
testing, and others noted that the campaign <i>“took
their data driven strategy to the next level.”</i> (Atlantic)<br />
<br />
<i>Microlistening:</i>The Obama campaign used text mining to analyze free form text input by volunteers to identify likelihood to donate, whether they would volunteer, and their perspective on political issues. Although many politicians have used rudimentary keyword search to monitor voter sentiment before, I am optimistic that improvements in free form text mining will change the interaction between voters and their representatives.<br />
<br />
<i>Crowdsourcing Political Activism:</i> Dashboard and Call Tool are crowdsourced
political activism. The phrase just sounds funny because political activism has
always been ‘crowdsourced’… just not digitally. I am confident this will become
a mainstay of world politics, and that it bodes well for democracy.
<br />
<br />
<div style="font-size: 6.0pt;">
<b><span style="font-size: 6.0pt;">Citations</span></b>
<br />
“Cambridge’s
BlueFin Labs decodes social media chatter” by Neil Swidey. The Boston Globe.
November 25, 2012.
<a href="http://www.bostonglobe.com/2012/11/25/cambridge-bluefin-labs-decodes-social-media-chatter/SLDp9nflJK0tFQKBPuVZhP/story.html?s_campaign=sm_tw">http://www.bostonglobe.com/2012/11/25/cambridge-bluefin-labs-decodes-social-media-chatter/SLDp9nflJK0tFQKBPuVZhP/story.html?s_campaign=sm_tw</a>
<br />
<br />
“Pundit
forecasts all wrong, Silver Perfectly Right. Is Punditry dead?” Techcrunch.
07NOV12.
<a href="http://techcrunch.com/2012/11/07/pundit-forecasts-all-wrong-silver-perfectly-right-is-punditry-dead/">http://techcrunch.com/2012/11/07/pundit-forecasts-all-wrong-silver-perfectly-right-is-punditry-dead/</a>
<br />
<br />
“When
the Nerds Go Marching In.” 16NOV12. The Atlantic.
<a href="http://www.theatlantic.com/technology/archive/2012/11/when-the-nerds-go-marching-in/265325/?single_page=true">http://www.theatlantic.com/technology/archive/2012/11/when-the-nerds-go-marching-in/265325/?single_page=true</a>
<br />
<br />
“Project
DreamCatcher: How Cutting Edge Text Analytics Can Help the President.” By Sasha
Issenberg. Slate. 13JAN2012.
<a href="http://www.slate.com/articles/news_and_politics/victory_lab/2012/01/project_dreamcatcher_how_cutting_edge_text_analytics_can_help_the_obama_campaign_determine_voters_hopes_and_fears_.single.html">http://www.slate.com/articles/news_and_politics/victory_lab/2012/01/project_dreamcatcher_how_cutting_edge_text_analytics_can_help_the_obama_campaign_determine_voters_hopes_and_fears_.single.html</a>
<br />
<br />
“Nate
Silver’s Political Calculus.”
<a href="http://fivethirtyeight.blogs.nytimes.com/">http://fivethirtyeight.blogs.nytimes.com/</a>
<br />
<a href="http://twitter.com/fivethirtyeight">http://twitter.com/fivethirtyeight</a>
<br />
<br />
“Built to
win: Deep inside Obama’s campaign tech.” By Sean Gallagher. November 14, 2012.
<a href="http://arstechnica.com/information-technology/2012/11/built-to-win-deep-inside-obamas-campaign-tech/">http://arstechnica.com/information-technology/2012/11/built-to-win-deep-inside-obamas-campaign-tech/</a>
<br />
<br />
“Consensus is
boring! Washington still won’t pay big bucks for predictions.” By Nick Arnett.
November 8<sup>th</sup>, 2012. VentureBeat.
<a href="http://venturebeat.com/2012/11/08/big-data-election-silver/">http://venturebeat.com/2012/11/08/big-data-election-silver/</a>
<br />
<br />
“Inside the
Secret World of Data Crunchers Who Helped Obama Win.” Time magazine. November 7, 2012.
<a href="http://swampland.time.com/2012/11/07/inside-the-secret-world-of-quants-and-data-crunchers-who-helped-obama-win/">http://swampland.time.com/2012/11/07/inside-the-secret-world-of-quants-and-data-crunchers-who-helped-obama-win/</a>
<br />
<br />
“Our
MoneyBall Election.” By Adam Gopnik. The New Yorker. 06NOV12.
<a href="http://www.newyorker.com/online/blogs/newsdesk/2012/11/our-money-ball-election.html">http://www.newyorker.com/online/blogs/newsdesk/2012/11/our-money-ball-election.html</a>
<br />
<br />
“The A/B
Test: Inside the Technology That’s Changing the Rules of Business. ” By Brian
Christian. Wired magazine. April 25, 2012.
<a href="http://www.wired.com/business/2012/04/ff_abtesting/all/">http://www.wired.com/business/2012/04/ff_abtesting/all/</a>
<br />
<br />
“Obama Uses Houdini.” The Daily Kos. 02NOV2008.
<a href="http://www.dailykos.com/story/2008/11/02/650137/-Obama-uses-Houdini">http://www.dailykos.com/story/2008/11/02/650137/-Obama-uses-Houdini</a>
<br />
<br />
Obama
Campaign Introduction to Dashboard video. <a href="http://www.youtube.com/watch?v=lqSGFnOwQkM">http://www.youtube.com/watch?v=lqSGFnOwQkM</a>
</div>
</div>CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.comtag:blogger.com,1999:blog-4323599490342956532.post-27244375537299637492012-11-13T23:23:00.002-05:002012-11-13T23:23:58.655-05:00Australian Olympic Success: Winning with Quants not Jocks<div dir="ltr" style="text-align: left;" trbidi="on">
<br />
<div class="MsoNormal" style="background: white; line-height: 15.0pt; margin-bottom: .0001pt; margin-bottom: 0in;">
<a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiPVPgY9qx0CXNT0eX-TvV-F_CW6EpwJ5E_X4jIsTVmFjErK6YCZPwLz1BtulzH32XshyphenhyphenfCeaXzw1BxZnzFOmXrnKmMfaxQOpS-72fnDBGo82xmIFV2dgGL7FziPW3Wt2GrqsW0wsRDQha9/s1600/Per+Capita+Medal+Count.JPG" imageanchor="1" style="clear: left; display: inline !important; float: left; line-height: 15pt; margin-bottom: 1em; margin-right: 1em;">
<img border="0" height="400" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiPVPgY9qx0CXNT0eX-TvV-F_CW6EpwJ5E_X4jIsTVmFjErK6YCZPwLz1BtulzH32XshyphenhyphenfCeaXzw1BxZnzFOmXrnKmMfaxQOpS-72fnDBGo82xmIFV2dgGL7FziPW3Wt2GrqsW0wsRDQha9/s400/Per+Capita+Medal+Count.JPG" width="170" /></a>
<br />
The July issue of Wired magazine gave a powerful example of Competitive
Advantage via Quantitative Methods in the Olympics. It pointed out that
Australia generates 6 times the medals per capita as the United States and four
times as many as any other members of the G20. The article also points out that
this is not a coincidence, but instead the rigorous application of the
scientific method to engineer exceptional athletic performance. The exponential
growth in Australian medal counts over the past 40 years clearly distinguishes their
paradigm as a competitive advantage.<br />
<br />
<a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjYd4P4zUS_SO9wrgl58Um0G97QP7aSDuWxQWIVZI1BoEObV_jZXQp3SKpgaNYjHHgid_by6FkP_OlfTrhgWs6eUZCtpSKeRziUdgt_FjCcsDNMo1j14fj_WbyHe7TaKqJ3rVxQFKutCIM2/s1600/Australia+Medal+Trending.JPG" imageanchor="1" style="clear: right; float: center; margin-bottom: 1em; margin-left: 1em;">
<img border="0" height="185" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjYd4P4zUS_SO9wrgl58Um0G97QP7aSDuWxQWIVZI1BoEObV_jZXQp3SKpgaNYjHHgid_by6FkP_OlfTrhgWs6eUZCtpSKeRziUdgt_FjCcsDNMo1j14fj_WbyHe7TaKqJ3rVxQFKutCIM2/s320/Australia+Medal+Trending.JPG" width="320" />
</a>
</div>
<div class="MsoNormal" style="background: white; line-height: 15.0pt; margin-bottom: .0001pt; margin-bottom: 0in;">
Australian success is owed to the creation of the AIS, the Australian
Institute of Sport. The <a href="http://www.ausport.gov.au/ais/">AIS</a> is <i>“a global leader in merging science with athletic
talent.”</i> (McClusky) It was founded in 1976, in the model of Eastern Bloc athletics programs to make Australia shine at the Olympics. <i>“AIS hoped to capture the intensity and success of the Soviet academies, without going to the same excesses. The idea was simple: Get the best coaches and the best athletes together on a year-round basis, without any distractions, and hope that athletic magic would result.”</i>
(McClusky) It began with a focus on dedicated funding for year-round sponsorship of high potential athletes, but in 2000 expanded into broad-based
research focused on the creation of competitive advantage through nutrition, potential identification, performance quantification, recovery optimization and training optimization.
</div>
<br />
<div class="MsoNormal" style="background: white; line-height: 15.0pt; margin-bottom: .0001pt; margin-bottom: 0in;">
<b>Methods</b>
<br />
<ul>
<li><b>Quantification:</b>
<i>“A lot of things you don’t know, simply because you can’t
measure them.” “Getting data like this puts you in a position to ask
intelligent questions.”</i> (McClusky)</li>
<ul>
<li><b><a href="http://www.youtube.com/watch?v=8lEQ6lWP2Vo">OptoJump</a>: </b>System for footfall tracking.</li>
<li><b><a href="http://www.zdnet.com/technology-one-of-the-good-sports-for-asc-1339285372/">Minimax:</a></b>
GPS and computer installed on crew boats, developed at AIS for <i>“instant motion analysis of [rowing] athletes during competition.”</i> </li>
<li><b>AIS Swimming Facility: </b>
Includes<i> <b>“</b>30 cameras mounted above, around, and
under the water and a motorized cart that runs alongside the swimmers to capture
data on their strokes. One of the starting blocks is rigged with force plates
and motion sensors; a snake of bundled cables runs over the wet deck to a
computer with a huge plasma screen for a monitor.” </i>(McClusky)
<i>“The system maps all of the parameters of the start she’s
just performed—the amount of force she pushed with, the angle at which it was
applied, her angle of flight through the air, the distance she went before
entering the water, the angle of her entry, and her depth under the water.”</i>
(McClusky)</li>
<li><b>CSIRO Sailing Wind Forecasts: </b>
Scientists from CSIRO Marine and Atmospheric Research provided
the Australian sailing team with <a href="http://www.csiro.au/Portals/Media/CSIRO-technology-gives-Australian-Olympic-sailors-the-winning-edge.aspx">hyperaccurate near-real time wind forecasts</a> and contributing to Australia's 2012 Gold medal in sailing.</li>
</ul>
<li><b> AIS Funded Research: </b>
Much of this research is available on the AIS sponsor’s website, with one
research synopses stating <i>potential areas where a competitive advantage may be gained through the use of…”</i> clearly identifying this as CAvQM.</li>
<ul>
<li><a href="http://www.pponline.co.uk/encyc/altitude-training-effects.html">Benefits of altitude training</a> for aerobic athletes</li>
<li><a href="http://www.ausport.gov.au/sportscoachmag/sports_sciences/compression_garments_do_they_influence_athletic_performance_and_recovery">Benefits of compression garments</a> for recovery</li>
<li>Benefits of hydrotherapy for recovery</li>
<li>Developed a test for EPO, a performance enhancing drug</li>
<li>Determined that consuming beet juice <i>"can improve aerobic exercise performance by as much as 2 percent.”</i> (McClusky)</li>
</ul>
<li><b>Athlete Identification: </b>
It is more efficient to identify talent amongst professional
athletes than to let athletes pick the sport of their choice. China and the USSR would test children and invest heavily in their training, but even this practice is inefficient. It is much more effective to identify the key predictors of exceptional performance and
test large groups of professionals to identify the best performers. <i>
<ul>
<li>“They pioneered programs to identify athletic potential and even to find athletes who might excel at a different sport better suited to their abilities. “</li>
</ul>
</i>(McClusky)</li>
<li><i>“An Australian racer qualified for the
Olympics just 18 months after she first saw a [skeleton] sled. Amazingly, she had
completed only 220 runs before qualifying. (A typical US skeleton racer makes
upwards of 2,000 runs before appearing in the Olympics.)” </i>(McClusky)</li>
<li><i style="line-height: 15pt;">They determined that one significant predictor of success had nothing to do with the sled itself or even the skill of the pilot. The faster a competitor pushed the sled through the 30-meter start zone before jumping on it, the better they performed. So researchers set up a national testing campaign, looking for women with
backgrounds in competitive sports who excelled at the 30-meter sprint."</i><span style="line-height: 15pt;">
(McClusky)</span></li>
<li>AIS has even evaluated considered <a href="http://www.ausport.gov.au/sportscoachmag/sports_sciences/gene_testing_a_valuble_proposition">genomic evaluation of athletes</a></li>
</ul>
</div>
<div class="MsoNormal" style="background: white; line-height: 15.0pt; margin-bottom: .0001pt; margin-bottom: 0in;">
<br />
<b>Personal Commentary</b><br />
Some of these concepts such as athlete identification or quantification
of performance we’ve touched on in prior blogs.
This case differs in composition though, regarding the focus on communication and the lack of co-located technicians and athletes. It is only fleetingly mentioned in one article, but athletes are spread throughout the
country so AIS uses <i>“an extensive WAN infrastructure that enables access… through
the use of a Citrix WAN scaler and presentation server.”</i><span style="line-height: 15pt;"> (Browne) I've cited the importance of communications in creating competitive advantage, but it is nonetheless surprising to notice it in the athletic arena as well.</span></div>
<div class="MsoNormal" style="background: white; line-height: 15.0pt; margin-bottom: .0001pt; margin-bottom: 0in;">
<br /></div>
<div class="MsoNormal">
The ultimate irony may be though, that in developing a
research based competitive advantage in a free country, it is difficult if not impossible to prevent the export of that advantage to other countries. Accordingly, we must realize that <u>only research that can be protected as trade secrets can create enduring competitive advantage (by making the research's export an act of intellectual property theft).</u> </div>
<br />
<br />
<br />
<br />
<br />
<b>
<span style="font-size: xx-small;">Citations</span></b><br />
<div class="MsoNormal">
<span style="font-size: xx-small;"><br /></span></div>
<div class="MsoNormal">
<span style="font-size: xx-small;">“One Hundredth of a Second Faster: Building Better Olympic
Athletes.” By Marc McClusky. Wired. July 25, 2012.</span></div>
<div class="MsoNormal">
<span style="font-size: xx-small;"><a href="http://www.wired.com/playbook/2012/06/ff_superhumans/">http://www.wired.com/playbook/2012/06/ff_superhumans/</a></span></div>
<div class="MsoNormal">
<br /></div>
<div class="MsoNormal">
<span style="font-size: xx-small;">OptoJump Sales Materials on Youtube
</span></div>
<div class="MsoNormal">
<span style="font-size: xx-small;"><a href="http://www.youtube.com/watch?v=8lEQ6lWP2Vo">http://www.youtube.com/watch?v=8lEQ6lWP2Vo</a></span></div>
<div class="MsoNormal">
<br /></div>
<div class="MsoNormal">
<span style="font-size: xx-small;">“Technology One of the Good Sports for ASC.” By Marcus
Browne. ZDNet. January 25, 2008.</span></div>
<div class="MsoNormal">
<span style="font-size: xx-small;"><a href="http://www.zdnet.com/technology-one-of-the-good-sports-for-asc-1339285372/">http://www.zdnet.com/technology-one-of-the-good-sports-for-asc-1339285372/</a></span></div>
<div class="MsoNormal">
<br /></div>
<div class="MsoNormal">
<span style="font-size: xx-small;">“Compression garments: Do they influence athletic
performance and recovery?” By <em><span style="border: 1pt none windowtext; color: #777777; font-family: Arial, sans-serif; line-height: 115%; padding: 0in;">Lee Wallace, Katie Slattery and
Aaron Coutts, School of Leisure, Sport and Tourism, University of Technology,
Sydney</span></em><span class="apple-converted-space"><i><span style="border: 1pt none windowtext; color: #777777; font-family: Arial, sans-serif; line-height: 115%; padding: 0in;">. </span></i></span></span></div>
<div class="MsoNormal">
<span style="font-size: xx-small;"><a href="http://www.ausport.gov.au/sportscoachmag/sports_sciences/compression_garments_do_they_influence_athletic_performance_and_recovery">http://www.ausport.gov.au/sportscoachmag/sports_sciences/compression_garments_do_they_influence_athletic_performance_and_recovery</a></span></div>
<div class="MsoNormal">
<br /></div>
<div class="MsoNormal">
<span style="font-size: xx-small;">“Gene testing: a valuable proposition.” By Grant Nahill.</span></div>
<div class="MsoNormal">
<span style="font-size: xx-small;"><a href="http://www.ausport.gov.au/sportscoachmag/sports_sciences/gene_testing_a_valuble_proposition">http://www.ausport.gov.au/sportscoachmag/sports_sciences/gene_testing_a_valuble_proposition</a></span></div>
<div class="MsoNormal">
<br /></div>
<div class="MsoNormal">
<span style="font-size: xx-small;">“CSIRO technology gives Australian Olympic Sailors the
winning edge.” August 16, 2012.</span></div>
<div class="MsoNormal">
<span style="font-family: Arial, sans-serif; line-height: 115%;"><span style="font-size: xx-small;"><a href="http://www.csiro.au/Portals/Media/CSIRO-technology-gives-Australian-Olympic-sailors-the-winning-edge.aspx">http://www.csiro.au/Portals/Media/CSIRO-technology-gives-Australian-Olympic-sailors-the-winning-edge.aspx</a></span></span></div>
<div class="MsoNormal">
<br /></div>
</div>
CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.comtag:blogger.com,1999:blog-4323599490342956532.post-7997458103954196052012-08-29T01:34:00.002-04:002012-08-29T01:34:50.709-04:00American Airlines' and Delta’s Revenue Management Systems <div dir="ltr" style="text-align: left;" trbidi="on">
<b><span style="font-size: large;">Background</span></b><br />
<div class="MsoNormal">
</div>
<div class="MsoNormal">
Although I’ve referenced Harrah’s <a href="http://en.wikipedia.org/wiki/Yield_management">yield management</a> system in a
prior <a href="http://cavqm.blogspot.com/2010/08/harrahs-loyalty-system.html">blog</a>,
the first yield management system was developed by American Airlines in the
1980’s with the support of the CEO <a href="http://en.wikipedia.org/wiki/Robert_Crandall">Robert Crandall</a>. As the Edelman Prize
committee <a href="http://www.airlinerevenuemanagement.com/">would later note</a>, the system contributed <u>$1.4 billion in profit over three
years</u>. How could any computer system generate half a billion dollars a year, and what methods did it use to achieve such a feat? And given such enormous gains, can any modern airline survive without such capabilities?<o:p></o:p></div>
<div class="MsoNormal">
<br /></div>
<div class="MsoNormal">
<b><span style="font-size: large;">Methods & Philosophy of Revenue Management</span></b><br />
I was inspired to write this blog after reading <u>Revenue Management</u> by Cross. Given his wide-ranging <a href="http://en.wikipedia.org/wiki/Revenue_management">Revenue Management</a> ('RM') work and the detail he provides on RM in the airline and car rental industries, I tried to distill as much from that book as possible.</div>
<div class="MsoNormal">
<ul style="text-align: left;">
<li>Use <b>market-based pricing</b> rather than cost-based pricing.
</li>
<li><b>Use price to balance</b> supply and demand.</li>
<ul>
<li>Attract price sensitive customers to off-peak days and charge peak-day customers more.</li>
<li>Price to attract passengers on competitors' flights to empty seats on your flights.</li>
<li>Use innovative pricing strategies to earn some profit from <i>"perishable products and opportunities” </i>before spoilage.</li>
<li>Use price differentiation to smooth capital shortages resulting from <i>"seasonal demand peaks”.</i></li>
</ul>
<li><b>Microsegment your customers by every dimension possible.</b></li>
<ul>
<li><i>“Market segmentation is the key to market-based pricing and revenue maximization.”</i></li>
<li><i>“Prevent people willing to pay higher prices from paying lower ones.”</i></li>
<li>Segment markets, forecast demand, optimize response.</li>
<li>Segment customers by <i>“...their characteristics, including purchase patterns, perception of the product, and willingness to pay.”</i> (Cross)</li>
<li>An RM system must account for <i>“differences at the micromarket level, based on product type, day of week, and the individual location.”</i> (Cross)</li>
</ul>
<li><b>Forecasting</b></li>
<ul>
<li>Forecast demand at every possible pre-spoilage time and re-forecast frequently.</li>
<li>RM systems should use dynamic forecasting at the micromarket level.</li>
<li><i>“The demand forecasting piece would support length-of-rent optimization at the same time it forecast arrival date activity. The optimization function would address overbooking and apply minimum length-of-rent parameters in inventory recommendations.”</i></li>
<li><i>“Often, we consider the possibility of reforecasting after every customer transaction. Studies indicate that the improvements in decision making from such a reforecast can increase a company’s revenues 1%-2%.“</i> (Cross)</li>
<li>Deliberately withhold inventory during price wars to sell more seats at the undiscounted price as competitors run out of inventory early.</li>
</ul>
</ul>
</div>
<div class="MsoNormal">
</div>
<div class="MsoNormal">
<b><span style="font-size: large;">Quantification
of Benefits</span><o:p></o:p></b></div>
<div class="MsoNormal" style="margin-bottom: 0.0001pt;">
</div>
<ul style="text-align: left;">
<li><b>Significant
Revenue Driver: </b>As Bob Crandall indicated, <i>“We expect Yield Management to generate at
least $500 million annually for the foreseeable future.”</i></li>
<li><b>Significant
Revenue Driver: </b>Delta’s initial foray into RM produced $300 million in incremental revenue with 56 people, which represented 50% of
declared profits that year. (Cross)</li>
<li><b>Significant
Revenue Driver: </b>PanAm produced an incremental $70 million in
revenues from the implementation of low level RM capabilities. (Cross)</li>
<li><b>Incremental
Revenue:</b><i> “Revenue
gains of 3%-7% are often realized with relatively little incremental cost.”</i> (Cross)</li>
<li><i>“American
Airlines increased revenues 14.5% with 47.8% profit growth in 1985 despite a
pilot strike.”</i> (Cross)</li>
<li><b>Profitable
market share growth:</b> <i>“The RM
[advanced airlines] at the rollout of the Ultimate Super Save price war gained
15% of traffic and grew revenues 9% while the RM [laggard airlines] grew
traffic 18% with 2% revenue growth.”</i></li>
<li><b>ROI
Exceeding 200%: </b>RM systems can generate ROI three times higher
than that of the average IT system investment <a href="http://ebusiness.mit.edu/erik/Optimize/pr_roi.html">according to Erik Brynjofsson</a> of MIT.</li>
</ul>
<div>
<div class="MsoNormal">
<b><br /></b></div>
<div class="MsoNormal">
<b><span style="font-size: large;">Commentary</span><o:p></o:p></b></div>
<div class="MsoNormal">
The most important take away from airlines' experience with RM systems is that <u>RM IS microsegmentation</u>.
Secondly, <u>where inventory management and pricing are not managed jointly,
there will be opportunities for revenue improvement, profits and efficiency
enhancement</u>. Furthermore, RM systematizes pricing policies while optimizing
opportunity and creating instantaneous decision making speed. It removes human
bias from the equation, modularizes pricing knowledge, and facilitates employee
training. It allows microsegmentation on hundreds of dimensions more than is possible with human pricing, and lets managers focus on customer service rather
than pricing concerns.</div>
<div class="MsoNormal">
<o:p></o:p></div>
<div class="MsoNormal">
<br /></div>
<div class="MsoNormal">
It’s also enlightening that Wall Street penalized American
Airlines when they implemented RM and began to
steal market share from discount airlines. AMR stock fell more than 15% after
advertising their new prices, which investors thought was the beginning of a
‘suicidal’ price war. American remained silent about what they were doing, and
analysts were baffled when their quarterly earnings announcement showed
spectacular revenue growth.</div>
<div class="MsoNormal">
<b><br /></b></div>
<div class="MsoNormal">
</div>
<div class="MsoNormal" style="margin-bottom: 0.0001pt;">
“THE SEVEN CORE CONCEPTS OF REVENUE
MANAGEMENT:" Sell to microsegments. Reforecast
often. Use fact based decision making. Use market-based pricing and not
cost-based pricing. Use price rather than capital to balance supply and demand.
Save product for your most valuable customers. <i>“Exploit each product’s value cycle.”</i></div>
<div class="MsoNormal" style="margin-bottom: 0.0001pt;">
<o:p></o:p></div>
<div class="MsoNormal" style="margin-bottom: 0.0001pt;">
<br /></div>
<div class="MsoNormal" style="margin-bottom: 0.0001pt;">
“THE SEVEN UNCERTAINTIES:" Seasonality, spoilage, competitive pricing, market uncertainty,
bundling/wholesaling, demand by market segment, opportunity perishability.</div>
<div class="MsoNormal" style="margin-bottom: 0.0001pt;">
<o:p></o:p></div>
<br />
<div class="MsoNormal">
<b><br /></b></div>
<div class="MsoNormal">
<b><br /></b></div>
<div class="MsoNormal">
<b><span style="font-size: x-small;">Citations</span></b></div>
<div class="MsoNormal">
<u><span style="font-size: x-small;"><br /></span></u></div>
<div class="MsoNormal">
<span style="font-size: x-small;"><u><a href="http://www.revenueanalytics.com/insights_book.asp">Revenue Management</a>.</u> By <a href="http://www.revenueanalytics.com/about_executives.asp">RobertCross</a>. Published 1997, New York NY by Bantam Doubleday Dell Publishing Group.</span></div>
<div class="MsoNormal">
<span style="font-size: x-small;"><br /></span></div>
<div class="MsoNormal">
<a href="http://www.airlinerevenuemanagement.com/"><span style="font-size: x-small;">www.airlinerevenuemanagement.com</span></a></div>
<div class="MsoNormal">
<span style="font-size: x-small;"><br /></span></div>
<div class="MsoNormal">
<span style="font-size: x-small;">“ROI Valuation: The IT Productivity
Gap.” By Erik Brynjolfsson. July 2003.</span></div>
<div class="MsoNormal">
<a href="http://ebusiness.mit.edu/erik/Optimize/pr_roi.html"><span style="font-size: x-small;">http://ebusiness.mit.edu/erik/Optimize/pr_roi.html</span></a></div>
<div class="MsoNormal">
</div>
<div class="MsoNormal" style="margin-left: .25in;">
<o:p></o:p></div>
</div>
<o:p></o:p><br />
<div class="MsoNormal">
<o:p></o:p></div>
<div class="MsoNormal" style="margin-bottom: 0.0001pt;">
<o:p></o:p></div>
<div class="MsoNormal" style="margin-bottom: 0.0001pt;">
<o:p></o:p></div>
<div class="MsoNormal" style="margin-bottom: 0.0001pt;">
<o:p></o:p></div>
<div class="MsoNormal" style="margin-bottom: 0.0001pt;">
<o:p></o:p></div>
<div class="MsoNormal">
<o:p></o:p></div>
<br />
<br /></div>
CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.comtag:blogger.com,1999:blog-4323599490342956532.post-68385087350128387182012-07-27T22:00:00.000-04:002012-07-28T21:37:21.292-04:00A Data Driven Strategy for Online Dating Sites<div dir="ltr" style="text-align: left;" trbidi="on">
<span style="background-color: white;">Given the popularity of my prior<a href="http://cavqm.blogspot.com/2011/01/eharmony-integrating-intelligence-with.html"> blog </a>on <span id="goog_1836823640"></span>eHarmony<span id="goog_1836823641"></span> strategic advantage, I thought I'd delve back into the mining of dating site information. While this topic continues to appear in behavioral economics books (for example <a href="http://danariely.com/">Dan Ariely</a>'s <u><a href="http://boingboing.net/2010/06/17/arielys-upside-of-ir.html">Upside of Irrationality</a></u>), it seems that factual information becomes more rare as insights are monetized. This is no where more obvious than in <a href="http://match.com/">Match.com</a>'s 2011 purchase of <a href="http://www.okcupid.com/">OKCupid</a>, which was regularly posting data driven insights publicly (many of which are cited below). </span><br />
<br />
<b><span style="font-size: large;">Academic Research on Dating Site Interaction</span></b><br />
First I'll provide a collection of dating site facts from a wide array of sources, followed by excerpts from OKCupid's OKTrends blog (which has the richest insights by far).<br />
<ul style="text-align: left;">
<li><a href="http://blog.neu.com/2010/09/09/valuable-online-dating-statistics-7-facts-you-need-to-know-before-looking-for-love-online/" style="background-color: white;">Including a Photo In Your Profile Doubles Your Response Rate</a></li>
<li><a href="http://eharmonycracked.blogspot.com/2007/10/time-to-abuse-some-eharmony-statistics.html" style="background-color: white;"><i>"Fewer than 10% of users log in every day, so patience is key."</i></a></li>
<li>On average users are two inches shorter than they say in their profile.</li>
<li>On average users make 20% less than they say in their profile.</li>
<li><a href="http://www.nytimes.com/2011/11/13/fashion/online-dating-as-scientific-research.html?pagewanted=all"><i>"many daters would rather admit to being fat than liberal or conservative"</i></a></li>
<li><a href="http://www.nytimes.com/2011/11/13/fashion/online-dating-as-scientific-research.html?pagewanted=all"><i>"white people are reluctant to date outside their race"</i></a></li>
<li><i>"<span style="background-color: white;">Among face-to-face participants, <a href="http://www.blogger.com/goog_1241512276">liking was primarily </a></span><span style="background-color: white;"><a href="http://people.ischool.berkeley.edu/~atf/papers/shaw_thesis.pdf">associated with how much they thought their partners disclosed.</a> This was not the case for </span><span style="background-color: white;">the online participants, for whom liking was more related to their own self-disclosure."</span></i>
</li>
</ul>
<span style="font-size: large;"><span style="background-color: white;"><b>Dating Research from OkCupid:</b></span><i style="background-color: white;"><b> </b><a href="http://blog.okcupid.com/index.php/online-dating-advice-exactly-what-to-say-in-a-first-message/">"Exactly What to say in a First Message"</a></i></span><br />
<ol style="text-align: left;">
<li><i>Be literate.</i> (Using the words 'ur' 'u' 'wont' 'cant' 'realy' 'luv' 'wat' drastically reduces responses.)</li>
<li><i>Avoid physical compliments. </i>(Using the words 'sexy' 'beautiful' 'hot' 'cutie' reduce responses.)</li>
<li><i>Use an unusual greeting. </i>(Using 'hi' 'hey' 'hello' reduce response rates.</li>
<li><i>Bring up specific interests.</i> (Words that indicate interest in the other's interest increase response rates.)</li>
<li><i>If you're a guy, be self effacing.</i> (Using 'sorry' 'apologize' 'awkward' all increase response rates.)</li>
<li><i>Consider becoming an athiest.</i> (Using the word 'athiest' increase response rates but the word 'God' decreases it.)</li>
</ol>
<span style="font-size: large;"><span style="background-color: white;">Dating Research from OkCupid:</span><i style="background-color: white;"> <a href="http://blog.okcupid.com/index.php/the-4-big-myths-of-profile-pictures/">Profile Pictures</a></i></span><br />
<ul style="text-align: left;">
<li>Women with flirtatious faces in pictures get the most responses, but smiling pictures get nearly as many responses. Women get substantially more responses when pictures make eye contact with the camera.</li>
<li>Men get the most responses when their pictures are looking away from the camera and aren't smiling.</li>
<li>Women taking pictures from over their head using a camera (the "MySpace angle") get twice as many responses as from any other picture type.</li>
<li>Men get 50% more responses for pictures with animals, pictures without their shirt on (if in shape), or doing something interesting. Note: Pictures with shirts off decrease in efficacy between 19-30, with little affect on response rates at ages over 30.</li>
<li>Women showing cleavage see 24%-80% more responses, with the greater response rates occurring at higher ages.</li>
<li>Better responses (i.e. leading to extended communication) occur when pictures contain animals, involve the subject doing something interesting, or are travel photos.</li>
</ul>
<br />
<span style="font-size: large;"><b>Dating Research from OkCupid:</b><i> <a href="http://blog.okcupid.com/index.php/dont-be-ugly-by-accident/">Don't Be Ugly By Accident!</a></i></span><br />
<ul style="text-align: left;">
<li>Using a flash makes the subject look 7 years older.</li>
<li>People are most attracted to photos taken shortly after sunrise or shortly before sunset (probably because the yellow light is the best illumnination).</li>
<li>The Panasonic Micro 4/3 S takes the best photos (as judged by people rating photos).</li>
</ul>
<br />
<b><span style="font-size: large;">Commentary</span></b><br />
<a href="http://healthland.time.com/2011/01/03/will-facebook-steal-online-dating-sites-girl/">Facebook's will become a paid Dating Site</a>. It would be an elegant application of their data, their stock price is falling as they look for new revenue streams, and they hired Dr. Andrew Fiore 10 months ago (who's been researching online dating behavior for a decade). Plus, an Economist article pegged <a href="http://www.economist.com/node/17797424/print">dating industry revenues at $3-4 billion dollars</a>.<br />
<br />
"How Can Anyone Use This Information" you ask? Here are my suggestions:<br />
<br />
<ul style="text-align: left;">
<li>If you're male, add two inches to your height and 20% to your income. If you're female indicate a more favorable body type and reduce your claimed weight by 8.5 lbs.</li>
<li>Buy a Panasonic Micro 4/3 S camera and use it for all of your profile shots.</li>
<li>Deliberately get photographs in yellow light (dawn or dusk).</li>
<li>Never use flash in your photographs.</li>
<li>If you're male, post profile shots where you're not looking at the camera, with pets, and/or doing something interesting. If female, make eye contact with the camera, flirt with the camera and use the 'mySpace shot'.</li>
<li>Men should never wear extra clothes in pictures, and women should wear revealing clothing.</li>
<li>Grammar check and spell check your communications (I wonder why the sites don't do this for you?</li>
<li>Never compliment physical characteristics in communication.</li>
<li>Use 'how's it going?' as a greeting.</li>
</ul>
<br />
<br />
<b>Citations</b><br />
<span style="font-size: x-small;"><br /></span><br />
<span style="font-size: x-small;">"Love, Lies and What They Learned." By Stephanie Rosenbloom.</span><br />
<span style="font-size: x-small;">The New York Times. November 12, 2011</span><br />
<span style="font-size: x-small;"><a href="http://www.nytimes.com/2011/11/13/fashion/online-dating-as-scientific-research.html?pagewanted=all">http://www.nytimes.com/2011/11/13/fashion/online-dating-as-scientific-research.html?pagewanted=all</a>
</span><br />
<span style="font-size: x-small;"><br /></span><br />
<span style="font-size: x-small;">"Will Facebook Kill Online Dating Sites' Girl?" By Belinda Luscombe. Time magazine. January 3, 2011.</span><br />
<span style="font-size: x-small;"><a href="http://healthland.time.com/2011/01/03/will-facebook-steal-online-dating-sites-girl/">http://healthland.time.com/2011/01/03/will-facebook-steal-online-dating-sites-girl/</a>
</span><br />
<span style="font-size: x-small;"><br /></span><br />
<span style="font-size: x-small;">Online Dating: Love at first bite." The Economist. December 29, 2010.</span><br />
<a href="http://www.economist.com/node/17797424/print">http://www.economist.com/node/17797424/print</a>
<br />
<span style="font-size: x-small;"><br /></span><br />
<span style="font-size: x-small;">"Dating Research From OkCupid; Exactly What To Say In A First Communication." By Christian Rudder. September 19, 2009.</span><br />
<span style="font-size: x-small;"><a href="http://blog.okcupid.com/index.php/online-dating-advice-exactly-what-to-say-in-a-first-message/">http://blog.okcupid.com/index.php/online-dating-advice-exactly-what-to-say-in-a-first-message/</a>
</span><br />
<span style="font-size: x-small;"><br /></span><br />
<span style="font-size: x-small;">"Dating Research From OkCupid; The 4 Big Myths Of Profile Pictures." By Christian Rudder. January 20th, 2010.</span><br />
<span style="font-size: x-small;"><a href="http://blog.okcupid.com/index.php/the-4-big-myths-of-profile-pictures/">http://blog.okcupid.com/index.php/the-4-big-myths-of-profile-pictures/</a>
</span><br />
<span style="font-size: x-small;"><br /></span><br />
<span style="font-size: x-small;">"Dating Research From OkCupid Don't Be Ugly By Accident!" By Christian Rudder. August 10th, 2010.</span><br />
<span style="font-size: x-small;"><a href="http://blog.okcupid.com/index.php/dont-be-ugly-by-accident/">http://blog.okcupid.com/index.php/dont-be-ugly-by-accident/</a>
</span><br />
<span style="font-size: x-small;"><br /></span><br />
<span style="font-size: x-small;"><a href="http://eharmonycracked.blogspot.com/2007/10/time-to-abuse-some-eharmony-statistics.html">http://eharmonycracked.blogspot.com/2007/10/time-to-abuse-some-eharmony-statistics.html</a>
</span><br />
<span style="font-size: x-small;"><br /></span><br />
<span style="font-size: x-small;"><a href="http://www.boston.com/bostonglobe/ideas/articles/2010/08/22/data_mining_the_heart/">http://www.boston.com/bostonglobe/ideas/articles/2010/08/22/data_mining_the_heart/</a>
</span><br />
<span style="font-size: x-small;"><br /></span><br />
<span style="font-size: x-small;"><a href="http://www.fastcompany.com/1812010/online-dating-okcupid-perfectmatch-pepper-schwartz">http://www.fastcompany.com/1812010/online-dating-okcupid-perfectmatch-pepper-schwartz</a>
</span><br />
<span style="font-size: x-small;"><br /></span><br />
<br /></div>CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.comtag:blogger.com,1999:blog-4323599490342956532.post-29929665211655393592012-05-09T19:59:00.002-04:002012-05-09T19:59:16.732-04:00Data-Driven Decision Making (DDDM) as an Intangible Asset<div dir="ltr" style="text-align: left;" trbidi="on">
<br />
<div class="MsoNormal">
<span style="font-family: Georgia, serif;">A fascinating New York Times <a href="http://www.nytimes.com/2011/04/24/business/24unboxed.html">article</a> </span><span style="font-family: Georgia, serif;">on Data-Driven Decision Making (DDDM) got my attention recently.
Professors at MIT and Wharton <a href="http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1819486">concluded</a> </span><span style="background-color: white; font-family: Georgia, serif;">that companies using DDDM
</span><i style="background-color: white; font-family: Georgia, serif;">“achieved productivity that was 5 to 6
percent higher than could be explained by other factors” (Lohr). </i><span style="background-color: white; font-family: Georgia, serif;">This
prompted the question, “What caused this 6% productivity advantage?” Or perhaps
more importantly, why is there so little information on data-driven decision
making outside the field of education?</span></div>
<div class="MsoNormal" style="background-attachment: initial; background-clip: initial; background-color: white; background-image: initial; background-origin: initial; background-position: initial initial; background-repeat: initial initial;">
<br /></div>
<div class="MsoNormal" style="background-attachment: initial; background-clip: initial; background-color: white; background-image: initial; background-origin: initial; background-position: initial initial; background-repeat: initial initial;">
<b><span style="font-family: Georgia, serif;"></span></b></div>
<h1>
</h1>
<h1>
<b>What is Data-Driven Decision Making?</b></h1>
<b><o:p></o:p></b><br />
<div class="MsoNormal" style="background-attachment: initial; background-clip: initial; background-color: white; background-image: initial; background-origin: initial; background-position: initial initial; background-repeat: initial initial; margin-bottom: 12pt;">
<span style="font-family: Georgia, serif;">The absence of internet content
on the strategic implications of DDDM forced me to precisely define it. To the
MIT professor who authored the research, DDDM is when ‘<i>decisions [are] based mainly on ‘data and analysis’ [rather than] the
traditional management arts of ‘experience and intuition.’” </i>Or as this was
phrased in my <a href="http://cavqm.blogspot.com/2011/10/microsoft-exp-and-experimentation-roi.html">blog on experimentation</a> platforms at Microsoft: <i>“HiPPO vs DDDM”</i></span><span style="font-family: Georgia, serif;">. It is important to
emphasize though, that every company collects data and produces reports using
it. The DDDM difference is made in the application of that data, when it is
mined for actionable insights. DDDM <i>“was
defined not only by collecting data, but also by how it is used — or not — in
making crucial decisions, like whether to create a new product or service.”</i><o:p></o:p></span></div>
<div class="MsoNormal" style="background-attachment: initial; background-clip: initial; background-color: white; background-image: initial; background-origin: initial; background-position: initial initial; background-repeat: initial initial;">
<span style="font-family: Georgia, serif;">Now that we’ve defined DDDM
for this discussion, it is time to revisit our original question, “What causes
the 6% ROI advantage enjoyed by DDDM companies?”<o:p></o:p></span></div>
<div class="MsoNormal" style="background: white; line-height: 17.6pt;">
<br /></div>
<div class="MsoNormal" style="background: white; line-height: 17.6pt;">
<b><span style="font-family: Georgia, serif;"></span></b></div>
<h1>
<b>Where does this ROI benefit come from?</b><span style="font-family: Georgia, serif;"> </span></h1>
<o:p></o:p><br />
<div class="MsoNormal" style="background: white;">
<span style="font-family: Georgia, serif;">The academic study did an admirable job
to confirm that there is a causal relationship between DDDM and ROI rather than
merely a correlation, so I will focus instead on micro examples that contribute
to the ROI effect. The contributors I suspect are feedback loops,
Experimentation ROI, and the avoidance of bad decisions.<o:p></o:p></span></div>
<div class="MsoNormal" style="background: white; line-height: 17.6pt;">
<span style="font-family: Georgia, serif;"><br /></span></div>
<div class="MsoNormal" style="background-attachment: initial; background-clip: initial; background-color: white; background-image: initial; background-origin: initial; background-position: initial initial; background-repeat: initial initial;">
<span style="font-family: Georgia, serif;"><i>Feedback
Loops: </i></span><span style="font-family: Georgia, serif;">Wired magazine
published a fascinating <a href="http://www.wired.com/magazine/2011/06/ff_feedbackloop/all/1">article</a> </span><span style="font-family: Georgia, serif;">on feedback loops last year where they
stated, <i>“</i></span><i><span style="background-attachment: initial; background-clip: initial; background-color: white; background-image: initial; background-origin: initial; font-family: Georgia, serif;">feedback loops have been thoroughly researched and validated
in psychology, epidemiology, military strategy, environmental studies,
engineering, and economics.” </span></i><span style="background-attachment: initial; background-clip: initial; background-color: white; background-image: initial; background-origin: initial; font-family: Georgia, serif;">More impressively though, they
indicated that <i>“feedback loops more
typically improve outcomes by about 10 percent compared to traditional methods.
That 10 percent figure is surprisingly persistent, it turns up in everything
from home energy monitors to smoking cessation programs to speed signs.” </i>It
seems that just giving people the information they need to monitor their
progress over time inspires them to improve that metric. Accordingly, we can
expect companies that monitor performance over time and make that information
accessible to employees would outperform companies that do not.</span></div>
<div class="MsoNormal" style="background-attachment: initial; background-clip: initial; background-color: white; background-image: initial; background-origin: initial; background-position: initial initial; background-repeat: initial initial;">
<span style="background-attachment: initial; background-clip: initial; background-color: white; background-image: initial; background-origin: initial; font-family: Georgia, serif;"><br /></span></div>
<div class="MsoNormal" style="background-attachment: initial; background-clip: initial; background-color: white; background-image: initial; background-origin: initial; background-position: initial initial; background-repeat: initial initial;">
<span style="font-family: Georgia, serif;"><i>Experimentation
ROI: </i></span><span style="font-family: Georgia, serif;">As I mentioned in a <a href="http://cavqm.blogspot.com/2011/10/microsoft-exp-and-experimentation-roi.html">prior blog</a>, Microsoft earned unbelievable ROI (i.e. 1,000,000%) on many
experimentation projects, saving millions of dollars in the process. The
projects required minimal investment and technology, with only three headcount
involved. Similarly, GoldCorp earned a 1,000,000% ROI on the GoldCorp
Challenge, identifying billions of dollars in rich deposits through a $675,000
prize. With such ROI possible, experimentation alone could account for the
productivity difference between DDDM and non-DDDM companies.<b><o:p></o:p></b></span></div>
<div class="MsoNormal" style="background-attachment: initial; background-clip: initial; background-color: white; background-image: initial; background-origin: initial; background-position: initial initial; background-repeat: initial initial;">
<span style="font-family: Georgia, serif;"><br /></span></div>
<div class="MsoNormal" style="background-attachment: initial; background-clip: initial; background-color: white; background-image: initial; background-origin: initial; background-position: initial initial; background-repeat: initial initial;">
<span style="font-family: Georgia, serif;"><i>Bias
Avoidance, Democratized Decision Making, Bad Decision Avoidance:</i> </span><span style="font-family: Georgia, serif;">Also mentioned in my Microsoft
Experimentation blog is their conclusion that 80% of business decisions based
on intuition are wrong. Given that experimentation, testing, and many other
objective means of decision evaluation avoid this 80% of errors, I expect this
‘bad decision avoidance’ to greatly impact productivity metrics. Every bad
decision avoided translates into greater employee productivity (just because an
employee is busy doesn’t mean they are creating value), greater revenue (as
more profitable projects are pursued), and resources are reallocated to more
profitable endeavors.<b><o:p></o:p></b></span></div>
<div class="MsoNormal" style="background: white; line-height: 17.6pt;">
<br /></div>
<div class="MsoNormal" style="background: white; line-height: 17.6pt;">
<b><span style="font-family: Georgia, serif;"></span></b></div>
<h1>
<b>Creating a DDDM Culture</b></h1>
<b><o:p></o:p></b><br />
<div class="MsoNormal" style="margin-bottom: 0.0001pt;">
<span style="font-family: Georgia, serif;">I could find very
little information on how to create a DDDM culture but
I’ve collected a number of best practice ‘guesses’ below. <o:p></o:p></span></div>
<div class="MsoNormal" style="margin-bottom: 0.0001pt;">
<br /></div>
<div class="MsoNormal" style="margin-bottom: 0.0001pt;">
<i><span style="font-family: Georgia, serif;">Analytical
Executives: </span></i><span style="font-family: Georgia, serif;">As suggested in <u><a href="http://www.tomdavenport.com/books.html">Competing On Analytics</a></u> (which I
reiterated in a <a href="http://cavqm.blogspot.com/2012/01/do-successful-analytics-companies.html">blog</a> on Analytical CEOs)</span><span style="font-family: Georgia, serif;">, analytical companies usually have a quant for
a CEO. Given that most of these companies known as DDDM cultures (i.e. Google
testing 43 shades of blue to decide the optimal shade for a webpage), I
hypothesize that an analytical CEO helps to create a DDDM culture.<o:p></o:p></span></div>
<div class="MsoNormal" style="margin-bottom: 0.0001pt;">
<span style="font-family: Georgia, serif;"><br /></span><i><span style="font-family: Georgia, serif;">Worker
Empowerment:</span></i><span style="font-family: Georgia, serif;"> The MIT research study cites other academic studies by
Bresnahan 2002 and Galal 1998 that identified <i>‘decentralized decision making’</i>
as a significant predictor of gains from DDDM. Avinash Kaushik also says
that analysts must have the autonomy and freedom to research what they think is
most important. He even goes so far as to suggest that analysts should have 20%
of their time to research questions that no one has asked them.</span></div>
<div class="MsoNormal" style="margin-bottom: 0.0001pt;">
<br /></div>
<div class="MsoNormal" style="background-attachment: initial; background-clip: initial; background-color: white; background-image: initial; background-origin: initial; background-position: initial initial; background-repeat: initial initial;">
<i><span style="font-family: Georgia, serif;">DDDM as DDM (Democratic Decision Making):</span></i><span style="font-family: Georgia, serif;">
DDDM is also the most <a href="http://www.curiousjuice.com/blog-0/bid/123216/Risks-of-data-driven-decision-making">democratic</a> decision making </span><span style="font-family: Georgia, serif;"> </span><span style="font-family: Georgia, serif;">approach
because it implies that anyone can have the best idea, and the only arbiter is
the data. This removes decision making as the province of the HiPPO (Highest
Paid Person’s Opinion), and recasts decision making as the province of the
customer. The <u>data</u> is the voice of the customer, whether in the form of surveys,
experiments, or testing and that’s the only voice that counts.<i> “Very few people, “HiPPO’s included, can
argue with a <a href="http://www.kaushik.net/avinash/seven-steps-to-creating-a-data-driven-decision-making-culture/">customer’s voice.</a>”</i></span><i><span style="font-family: Georgia, serif;"><o:p></o:p></span></i></div>
<div class="MsoNormal" style="background-attachment: initial; background-clip: initial; background-color: white; background-image: initial; background-origin: initial; background-position: initial initial; background-repeat: initial initial;">
<span style="font-family: Georgia, serif;"><i><br /></i></span></div>
<div class="MsoNormal">
<i><span style="font-family: Georgia, serif;">Reporting
is NOT Analysis: </span></i><span style="font-family: Georgia, serif;"> </span><span style="font-family: Georgia, serif;">Kaushik quips in another <a href="http://www.kaushik.net/avinash/refuse-report-requests-answer-analytics-business-questions/">article</a>, <i>“If
you need reporting, hire an intern.”</i> </span><span class="MsoHyperlink"><b><span style="font-family: Georgia, serif;"> </span></b></span><span style="font-family: Georgia, serif;">Analysts
should be asked difficult, non-straightforward business questions so that they
can learn the context of the question and answer it as thoroughly as their
domain expertise allows. <i>“Reward analysis and not the number of
emailed reports.”</i></span></div>
<div class="MsoNormal">
<span style="font-family: Georgia, serif;"><i><br /></i></span></div>
<div class="MsoNormal">
<i><span style="font-family: Georgia, serif;">Analytics Must Drive Outcomes: </span></i><span style="font-family: Georgia, serif;">When creating a DDDM culture, focus your
efforts on improving the outcomes that benefit your peers’ <u>compensation</u>
because it will quickly win you the support of your colleagues. Much more
important though, is that <u>analytics should be in a department with revenue
goals to which it can contribute</u>. This is the only way to truly focus
efforts on value adding activities. If analysts know that their ingenuity will
affect their group’s compensation, and thereby their own compensation, they
will search out opportunities like a bloodhound. <o:p></o:p></span></div>
<div class="MsoNormal">
<span style="font-family: Georgia, serif;"><br /></span></div>
<div class="MsoNormal" style="background: white; line-height: 17.6pt;">
<b><span style="font-family: Georgia, serif;"></span></b></div>
<h1>
<b>My Thoughts</b></h1>
<b><o:p></o:p></b><br />
<div style="background: white; margin-bottom: .0001pt; margin: 0in;">
<strong><span style="font-family: Georgia, serif; font-size: 11pt; font-weight: normal;">This blog
was a bit lengthy, so I only want to draw your attention to one sentence in the
research paper: </span></strong><i><span style="font-family: Georgia, serif;">“Collectively, our <u>results suggest that
DDDM capabilities can be modeled as intangible assets which are valued by
investors and which increase output and profitability</u>.”</span></i><span style="font-family: Georgia, serif;"> The last time I heard of a new class of
intangible assets, it was the Dot Com boom when every company was promoting
their first CTO/CIO. Perhaps we’re in the midst of the big data boom and we’ll soon see a
wave of Chief Data Officer promotions?<o:p></o:p></span></div>
<div class="MsoNormal">
<br /></div>
<div class="MsoNormal">
<br /></div>
<div class="MsoNormal">
<br /></div>
<div class="MsoNormal">
<span style="font-family: Georgia, serif;"><br /></span></div>
<div class="MsoNormal">
<b><span style="font-family: Georgia, serif;"></span></b></div>
<b>Citations</b>
<b><o:p></o:p></b><br />
<div class="MsoNormal">
<span style="font-family: Georgia, serif;"><small><o:p></o:p></small></span></div>
<small>
</small><br />
<div class="MsoNormal">
<small><span style="font-family: Georgia, serif;">“When There’s No
Such Thing As Too Much Information.” By Steve Lohr. The New York Times. April 23<sup>rd</sup>, 2011.<o:p></o:p></span></small></div>
<small>
<div class="MsoNormal">
<a href="http://www.nytimes.com/2011/04/24/business/24unboxed.html"><span style="font-family: Georgia, serif;">http://www.nytimes.com/2011/04/24/business/24unboxed.html</span></a><span style="font-family: Georgia, serif;"><o:p></o:p></span></div>
<div class="MsoNormal">
<br /></div>
<div class="MsoNormal">
<span style="font-family: Georgia, serif;">“Strength In
Numbers: How Does Data Driven Decisionmaking Affect Firm Performance?” By Erik Brynjolfsson, Lorin Hitt, Heekyung
Hellen Kim. April 22, 2011.<o:p></o:p></span></div>
<div class="MsoNormal">
<a href="http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1819486"><span style="font-family: Georgia, serif;">http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1819486</span></a><span class="MsoHyperlink"><span style="font-family: Georgia, serif;"><o:p></o:p></span></span></div>
<div class="MsoNormal">
<br /></div>
<div class="MsoNormal">
<span style="font-family: Georgia, serif;">“Risks of Data
Driven Decision Making.” By Arjun Moorthy. 08APR2012.<o:p></o:p></span></div>
<div class="MsoNormal">
<a href="http://www.curiousjuice.com/blog-0/bid/123216/Risks-of-data-driven-decision-making"><span style="font-family: Georgia, serif;">http://www.curiousjuice.com/blog-0/bid/123216/Risks-of-data-driven-decision-making</span></a><span style="font-family: Georgia, serif;"><o:p></o:p></span></div>
<div class="MsoNormal">
<br /></div>
<div class="MsoNormal">
<span style="font-family: Georgia, serif;">“Seven Steps To
Creating a Data Driven Decision Making Culture.” By Avenash Kaushik. 23OCT2006.<o:p></o:p></span></div>
<div style="background: white; line-height: 15.75pt;">
<a href="http://www.kaushik.net/avinash/seven-steps-to-creating-a-data-driven-decision-making-culture/"><span style="font-family: Georgia, serif; font-size: 11pt;">http://www.kaushik.net/avinash/seven-steps-to-creating-a-data-driven-decision-making-culture/</span></a><span style="font-family: Georgia, serif; font-size: 11pt;"><o:p></o:p></span></div>
<div class="MsoNormal">
<br /></div>
<div class="MsoNormal">
<span style="font-family: Georgia, serif;">“Rebel! Refuse
Reporting Requests. Only Answer Business Questions, FTW. “ Avenash Kaushik. 04OCT2010.<o:p></o:p></span></div>
<div class="MsoNormal">
<a href="http://www.kaushik.net/avinash/refuse-report-requests-answer-analytics-business-questions/"><span style="font-family: Georgia, serif;">http://www.kaushik.net/avinash/refuse-report-requests-answer-analytics-business-questions/</span></a><span style="font-family: Georgia, serif;"><o:p></o:p></span></div>
<div class="MsoNormal">
<br /></div>
<div class="MsoNormal">
<span style="font-family: Georgia, serif;">“Harnessing The
Power Of Feedback Loops.” By Thomas
Goetz. 19JUN2011. Wired. July 2011 Issue.<o:p></o:p></span></div>
<div class="MsoNormal">
<a href="http://www.wired.com/magazine/2011/06/ff_feedbackloop/all/1"><span style="font-family: Georgia, serif;">http://www.wired.com/magazine/2011/06/ff_feedbackloop/all/1</span></a><span style="font-family: Georgia, serif;"><o:p></o:p></span></div>
</small><br />
<div class="MsoNormal">
<small><span style="font-family: Georgia, serif;"></span></small><o:p></o:p></div>
</div>
CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.comtag:blogger.com,1999:blog-4323599490342956532.post-20704930790034254032012-03-30T20:29:00.001-04:002012-03-30T20:29:12.182-04:00Google's Mythical Merger Success Rate<div dir="ltr" style="text-align: left;" trbidi="on">
<h1>
Google Claims A 70% Acquisition Success Rate?</h1>
A recent Xconomy <a href="http://www.xconomy.com/san-francisco/2012/03/05/googles-rules-of-acquisition-how-to-be-an-android-not-an-aardvark/">article</a> about Google’s acquisition strategy grabbed my attention, because in it Google claimed a whopping <i>“70% success rate.”</i> According to a recent McKinsey <a href="http://www.xconomy.com/san-francisco/2012/03/05/googles-rules-of-acquisition-how-to-be-an-android-not-an-aardvark/">report</a> though, <i>“Anyone who has researched merger success rates knows that roughly 70% of mergers fail.”</i> Google's statement is less surprising when one examines their definition of acquisition failure from later in the article, <i>“The company’s definition of failure? When the acquired firm’s technology doesn’t get incorporated into a Google product and the company sunsets it, or the team eventually leaves, or both.”</i> Note that there is no mention of profitability or shareholder value creation. Accordingly, I think that Google’s 70% success rate is due more to their lower standard of success, than to process superiority.
<br />
<br />
Google may not worry about profits or creating shareholder value because their scarce resource is talent, and ‘acqui-hires’ are one way to accomplish that. Companies merge for a variety of reasons: expansion of market share, acquisition of new lines of distribution or technology, or reduction of operating costs. I suppose a criterion beyond shareholder value creation may be required for Google because they acquire so many startups, and the acquired companies are often too small to impact share price.
<br />
<br />
<h1>
Merger Success Statistics Summary</h1>
Even though I suspect that Google’s success rate is inflated, I still wanted to refresh my understanding of merger success. I remember being struck as an undergrad by a pie chart which showed that approximately a third of mergers destroyed value, a third didn't accomplish anything, and then of the value creating mergers only a portion of the value creating mergers did so significantly. I wasn't able to turn up that chart, but I did find the following statistics:<br />
<ul style="text-align: left;">
<li><b>70% Failure Rate: </b>McKinsey:<i>“Anyone who has researched merger success rates knows that roughly 70% of mergers fail.”</i></li>
<li><b>70% Failure Rate: </b>A 2004 <a href="http://www.bain.com/publications/articles/mega-merger-mouse-trap-wall-street-journal.aspx">study</a> by Bain & Company found that 70 percent of mergers failed to increase shareholder value.</li>
<li><b>90% Failure Rate in Europe: </b>A 2007 <a href="http://www.haygroup.com/ww/press/Details.aspx?ID=7276">study</a> by Hay Group and the Sorbonne found that more than 90 percent of mergers in Europe fail to reach financial goals.</li>
<li><b>Failure Exceeds 50%:</b> <i>"If the definition of a successful merger is driving up shareholder value, then their failure rate is far north of 50 percent,"</i> says Lawrence Chia, a managing director of Deloitte & Touche in Beijing, China.</li>
<li><b>Failure Exceeds 50%: </b>In Merrill DataSite’s <a href="http://www.datasite.com/cps/rde/xbcr/datasite/WP_PMI_mm_June_2009.pdf">survey</a> of 100 worldwide executives, the majority said that mergers are successful 26%-50% of the time in their experience. </li>
</ul>
<br />
<h1>
Reasons For Failure</h1>
Assuming that mergers are intended to increase shareholder value, there are many headwinds to overcome before shareholders benefit.<br />
<ul style="text-align: left;">
<li><b>Shareholder Dilution:</b> <i>“…when one company acquires another using its own stock as currency, as commonly happens today, shareholders' stakes in the acquiring firm typically decline.”</i></li>
<li><b>Cost of Golden Parachutes:</b> <i>“Company executives can pocket up to 8% of the merger proceeds”</i> according to Michael S. Kesner of Deloitte Consulting. </li>
<ul>
<li>Executive stock options vest immediately in the event of a merger. </li>
</ul>
<li><b>Inaccurate estimation of synergies:</b> The McKinsey report cites 32% of merged companies saying their synergy estimates were inaccurate. </li>
<li><b>Integration Risk</b></li>
<li><b>Bain Indicators of Failure: </b></li>
<ul>
<li>Billion Dollar Price Tag</li>
<li>Mergers of Equals</li>
<li>Hostile Mergers</li>
<li>Moves into a new line of business </li>
</ul>
</ul>
<br />
<h1>
Best Practices for Merger/Acquisition</h1>
<ul style="text-align: left;">
<li><b>Google:</b> Only buy companies that are <i>"directly aligned with your business."</i></li>
<li><i><b style="font-style: normal;">Bain:</b><span style="font-style: normal;"> Stay close to the core business.</span></i></li>
<li><b>Google:</b> Only buy a company <i>"if your resources will allow it to become something bigger."</i></li>
<li><b><b>IBM:</b><span style="font-weight: normal;"> Buy companies that will benefit from your distribution network and customer relationships. </span>
</b></li>
<li><b>Bain:</b> <i>"It turns out that frequent acquirers that build skills and experience through a host of small deals come out on top."</i></li>
<li><b>Bain:</b> <i>"The average deal size for the winners in our study was 10% of the acquiring firm's market cap."</i></li>
<li><b>Bain:</b> Only pursue friendly deals.</li>
<li><b>Bain:</b> Only pursue deals that are easy to justify, because it the benefit must be direct to succeed.</li>
<li>Buy companies that are significantly smaller so there is no question which culture will dominate.</li>
<li>Begin planning for integration while due diligence is occurring. </li>
<li>Quantify executive compensation during due diligence because it can dwarf synergies. </li>
<li>Be prepared to devote substantial executive effort to human capital and cultural issues.</li>
<li>Make headcount decisions quickly following the merger announcement to create certainty for employees. Otherwise, productivity will plummet.</li>
<li>Cross continent mergers are more likely to fail than domestic mergers.</li>
</ul>
<br />
<br />
<br />
<small>Citations</small>
<br />
<small><br /></small><br />
<small>
“What Are Mergers Good For?” By Gretchen Morganstern. The New York Times. June 5, 2005. </small><br />
<div>
<div>
<small><a href="http://www.nytimes.com/2005/06/05/magazine/05MERGERS.html?pagewanted=all">http://www.nytimes.com/2005/06/05/magazine/05MERGERS.html?pagewanted=all</a></small></div>
<div>
<small><br /></small></div>
<div>
<small> <a href="http://www.xconomy.com/san-francisco/2012/03/05/googles-rules-of-acquisition-how-to-be-an-android-not-an-aardvark/">http://www.xconomy.com/san-francisco/2012/03/05/googles-rules-of-acquisition-how-to-be-an-android-not-an-aardvark/ </a></small></div>
<div>
<small><br /></small></div>
<div>
<small> “Mergers Fail More Often Than Marriages.” By Kevin Voigt. CNN.com May 22, 2009.
<a href="http://edition.cnn.com/2009/BUSINESS/05/21/merger.marriage/index.html">http://edition.cnn.com/2009/BUSINESS/05/21/merger.marriage/index.html</a></small></div>
<div>
<small><br /></small></div>
<div>
<small>“Post Merger Integration: The Key To Successful Mergers and Acquisitions.” By Merrill Datasite. June 2009.
<a href="http://www.datasite.com/cps/rde/xbcr/datasite/WP_PMI_mm_June_2009.pdf">http://www.datasite.com/cps/rde/xbcr/datasite/WP_PMI_mm_June_2009.pdf </a></small></div>
<div>
<small><br /></small></div>
<div>
<small> “<a href="http://www.mckinsey.com/Client_Service/Organization/Latest_thinking/~/media/McKinsey/dotcom/client_service/Organization/PDFs/775084%20Merger%20Management%20Article%20Compendium.ashx">Beyond Risk Avoidance</a>: A McKinsey </small>Perspective on creating transformational value from mergers.” By James McCletchy and Andy West. <u><small>Perspectives on Merger Integration.</u> June 2010. Published by McKinsey and Company. Pg. 11.<br />
<br />
<a href="http://www.bain.com/publications/articles/mega-merger-mouse-trap-wall-street-journal.aspx">http://www.bain.com/publications/articles/mega-merger-mouse-trap-wall-street-journal.aspx</a>
<br />
<br />
<a href="http://www.haygroup.com/ww/press/Details.aspx?ID=7276">http://www.haygroup.com/ww/press/Details.aspx?ID=7276</a>
</small>
<br />
<div>
<div>
</div>
</div>
</div>
</div>
</div>CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.comtag:blogger.com,1999:blog-4323599490342956532.post-4342020418267071282012-03-14T22:24:00.000-04:002012-03-14T22:24:02.640-04:00Crowdsourcing Analytics: Restraints On Growth<div dir="ltr" style="text-align: left;" trbidi="on">
<br />
<div class="MsoNormal">
<b></b></div>
<h1>
<b>Is
Crowdsourced Analytics Effective?</b></h1>
<b><o:p></o:p></b><br />
<div class="MsoNormal">
I came across an interesting <a href="http://www.quora.com/The-Goldcorp-Challenge-who-has-replicated-it-successful">question</a> on Quora the other day about the relative lack of growth in crowdsourced
analytics over the past decade. The successes of the <a href="http://cavqm.blogspot.com/2012/02/goldcorp-challenge-and-beginning-of.html">GoldCorpChallenge</a> and the <a href="http://cavqm.blogspot.com/2011/05/1000000-netflix-challenge.html">NetflixChallenge</a> should have expanded the use of crowdsourced analytics competitions, so why hasn’t there been an explosion
of such competitions? Even the participants describe the competitions as a
goldmine for the analytical techniques developed, their use still hasn't taken off.<br />
<blockquote>
<i>
Although his team did not win,</i>
<i><span style="background-attachment: initial; background-clip: initial; background-color: white; background-image: initial; background-origin: initial;"> <a href="http://www.itbusinessedge.com/cm/blogs/all/crowdsourcing-can-help-companies-find-scarce-data-analysis-skills/?cs=46020%E2%80%9D%3E">the CEO ofdata analytics company Opera Solutions said</a> the company got "a
$10 million payoff internally from what we’ve learned” by using improved
modeling and analysis techniques it created for the contest with its paying
clients. And Netflix got a pretty sweet return on its investment. Companies
like Netflix could expect to pay $1 million to hire five researchers for a
year. As Darren Vengroff, a former lead researcher for Amazon's recommendation
engine, said in a<span class="apple-converted-space"> </span><strong><span style="font-family: Calibri, sans-serif;">Forbes</span></strong><span class="apple-converted-space"> </span>article, Netflix "spent the same
amount and got thousands, probably millions of engineer-years.</span></i></blockquote>
</div>
<div class="MsoNormal">
<br /></div>
<div class="MsoNormal">
<b></b></div>
<h1>
<b>Emergence
of Competitive Markets and Evidence of Growth</b></h1>
<b><o:p></o:p></b><br />
<div class="MsoNormal">
Some readers may point to the <a href="http://www.heritagehealthprize.com/c/hhp%E2%80%9D%3E">HeritageHealth Prize</a>’s $3 million
reward as evidence that crowdsourcing is growing, but it is only the third
multi-million dollar award offered in the past ten years. The following sites
now advertise themselves as marketplaces for crowdsourced talent and display
more than a hundred competitions in total, but crowdsourced analytics haven’t
caught fire as a discipline. Procter & Gamble uses Innocentive, <a href="http://informationweek.in/Software/12-01-30/How_an_Indian_startup_is_redefining_analytics_using_crowdsourcing.aspx">Kaggleis used by Ford and Microsoft</a>, so what is holding back the
marketplace?<o:p></o:p></div>
<div class="MsoListParagraphCxSpFirst" style="mso-list: l0 level1 lfo1; text-indent: -.25in;">
</div>
<ul style="text-align: left;">
<li><span style="font-family: Symbol; text-indent: -0.25in;"><span style="font-family: 'Times New Roman'; font-size: 7pt;"><a href="http://www.blogger.com/goog_1129572310"> </a></span></span><span style="color: windowtext; text-indent: -0.25in;"><a href="http://www.crowdanalytix.com/welcome">http://www.crowdanalytix.com/welcome</a></span></li>
<li><span style="color: windowtext; text-indent: -0.25in;"><a href="http://www.kaggle.com/">http://www.kaggle.com/</a></span></li>
<li><span class="MsoHyperlink" style="text-indent: -0.25in;"><span style="color: windowtext;"><a href="http://www.innocentive.com/">http://www.Innocentive.com/</a></span></span></li>
</ul>
<o:p></o:p><br />
<div class="MsoListParagraphCxSpLast" style="mso-list: l0 level1 lfo1; text-indent: -.25in;">
<o:p></o:p></div>
<div class="MsoListParagraphCxSpLast" style="mso-list: l0 level1 lfo1; text-indent: -.25in;">
<span class="MsoHyperlink"><span style="color: windowtext;"><br /></span></span></div>
<div class="MsoNormal">
<b></b><br />
<h1>
<b>Restraints
on the Growth of Crowdsourced Analytics</b></h1>
</div>
<b><o:p></o:p></b><br />
<div class="MsoNormal">
<i>Prestige and Marketing: </i>The winner of the Goldcorp Challenge earned
less in prize money than they spent to win the competition. They were more
interested in the free publicity it would generate for their startup business,
and wanted to use their victory as marketing collateral. Similarly, the runner
up in the Netflix Challenge used their success to launch Opera Analytics, a
consultancy focused on big data, which now leads the scoreboard in the Heritage
Health Prize. Such prestige only results from interesting, well funded
competitions with many participants which demands large prizes.</div>
<div class="MsoNormal">
<o:p></o:p></div>
<div align="center" class="MsoNormal" style="text-align: center;">
<b><br /></b></div>
<div align="center" class="MsoNormal" style="text-align: center;">
<b>To
generate interest from top talent, the prize amount must create lots of
publicity.<o:p></o:p></b></div>
<div class="MsoNormal">
<i><br /></i></div>
<div class="MsoNormal">
<i>Prize
Money and ROI:</i> <b>Companies are
attracted to crowdsourcing by the obscene ROI</b>, and they can
get acceptable improvements for much smaller price amounts. The prizes offered
on Kaggle and other sites are 95-99% smaller than the GoldCorp Challenge and
99% smaller than the Heritage Health Prize, ranging from $10,000-$30,000. At
this size value, participants aren’t attracted by the chance of winning unless
they’re living in emerging economies. <b>The prize amount determines whether the best data scientist in the world will partake or whether it will attract just the hobbyists.</b> Companies earn a better ROI on small
prize values, and may be content with the quality of results they receive so
effort on a grander scale is not required. <o:p></o:p></div>
<div class="MsoNormal">
<i><br /></i></div>
<div class="MsoNormal">
<i>Access to
Proprietary Data and Creation of Intellectual Property: </i>Like the
GoldCorp Challenge, the winners of the Netflix Challenge earned less in prize
money than they spent to win the competition. AT+T Labs dedicated three data scientists to the Netflix
Challenge for three years because the resulting intellectual property was
valuable to them and not because the project was profitable.<o:p></o:p></div>
<div class="MsoNormal">
<i><br /></i></div>
<div class="MsoNormal">
<i>Legal
Risks:</i> Netflix was ready to sponsor a second challenge, but reconsidered
after a class-action lawsuit began against them for disclosing customer information.
It was even more problematic after security researchers de-anonymized some
customer records by cross-referencing data from the Netflix Challenge with an
unaffiliated IMDB-like rating website.<o:p></o:p></div>
<div class="MsoNormal">
<i><br /></i></div>
<div class="MsoNormal">
<i>Most
Winners Are Corporate Sponsored: </i>The biggest competitions were won, or
appear they will be won by corporate-sponsored talent. If Opera Analytics
invests to win every notable competition as a marketing tactic, it discourages
non-corporate participation in the future.<o:p></o:p></div>
<div class="MsoNormal">
<i><br /></i></div>
<div class="MsoNormal">
<i>Winner’s
Curse:</i> The competition rules usually make the winner give a free license
to the sponsor (if not the patent). The runner up retains exclusive rights to
their intellectual property though. The question then is,<br />
<div style="text-align: center;">
<b>“Why not deliberately be the runner up to retain patent rights?”</b><o:p></o:p></div>
</div>
<div class="MsoNormal">
<b><br /></b></div>
<div class="MsoNormal">
<i>Competitive
Differentiation:</i> Since the competition is public and your
competitors could simply buy the analytical method from the participants
afterward, it doesn't create a competitive advantage. Accordingly, companies can’t differentiate themselves with crowdsourced
analytics. At best, a company can increase their ROI and remove
analytics from the arena as a competitive advantage. <b style="text-align: center;">Crowdsourced
analytics is great for destroying a competitors’ analytical
advantage, but doesn't necessarily create one.</b></div>
</div>CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.comtag:blogger.com,1999:blog-4323599490342956532.post-46201245782964784722012-02-22T23:51:00.000-05:002012-02-22T23:51:11.434-05:00The GoldCorp Challenge and the Beginning of Crowdsourced Analytics<div dir="ltr" style="text-align: left;" trbidi="on">
<h1>
The Beginning of Crowdsourced Analytics</h1>
Despite all of the press it got, the <a href="http://cavqm.blogspot.com/2011/05/1000000-netflix-challenge.html">Netflix Challenge</a> wasn’t the first
crowdsourced analytics problem. That
distinction goes to the GoldCorp Challenge, which was a last ditch
effort to save a gold mining company. Fortunately
for the company, ownership changed in a proxy fight and an industry
outsider ended up at the helm, Rob McEwen. In a true flash of genius, the new CEO crowdsourced a solution five years before Jeff Howe would coin the term in his famous Wired <a href="http://www.wired.com/wired/archive/14.06/crowds.html">article</a>. Even more intriguing, is that McEwen was inspired by the open source Linux development paradigm and asked, "How can I apply that to the mining industry?"<br />
<i><br /></i><br />
<h1>
The
Challenge</h1>
<b><o:p></o:p></b><br />
<i>"Rob McEwen triggered the gold rush by <a href="http://www.ideaconnection.com/open-innovation-success/Open-Innovation-Goldcorp-Challenge-00031.html">issuing an extraordinary challenge</a> to the world's geologists: We'll show you all of our data on the Red Lake mine online if you tell us where we're likely to find the next 6 million ounces of gold. The price: a total of $575,000, with a top award of $105,000." </i><i>"The external response was immediate. More than 1,400 scientists, engineers, and geologists from 50 countries downloaded the company's data and started their virtual exploration." </i>This opportunity did carry with it some risks though, such as a hostile take-over by a competitor using their data.<br />
<br />
<div class="separator" style="clear: both; text-align: center;">
<a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgbWCxc-bZwvD98TuuHe2Zmv5-b_hd_nf-pMkPdPulQIVanwKqE88wxWQTU_Za_4EjaPeCXg6FWEGlbcvw-MkeW5dO0HcWJL0kR9F6G_tKCn8kdRzM5iuq1TrNWdEF9IK5iBeag2aROoJUx/s1600/GoldCorp+Red+Lake+Winning+Visualization.gif" imageanchor="1" style="clear: right; float: right; margin-bottom: 1em; margin-left: 1em;"><img border="0" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgbWCxc-bZwvD98TuuHe2Zmv5-b_hd_nf-pMkPdPulQIVanwKqE88wxWQTU_Za_4EjaPeCXg6FWEGlbcvw-MkeW5dO0HcWJL0kR9F6G_tKCn8kdRzM5iuq1TrNWdEF9IK5iBeag2aROoJUx/s1600/GoldCorp+Red+Lake+Winning+Visualization.gif" /></a></div>
Fortune smiled on Goldcorp though, because in that group of 1,400 geologists was a high-powered Australian geological analysis firm,<a href="http://www.fractaltechnologies.com/case_study_goldcorp_challenge.25.html"> Fractal Analysis</a>, that was trying to make a name for itself. They ultimately won the competition, but conceded that they lost money on the analysis in hopes that the publicity would bring them more business. An image from their winning submission is included at the right. <i>"The worth of [the gold discovered] has so far exceeded $6 billion in value."</i><br />
<i><br /></i><br />
<h1>
<b>Quantification of Performance</b></h1>
<div class="MsoNormal">
<br />
<ul style="text-align: left;">
<li>Productivity and Cost
Improvements: Gold production increased from <a href="http://www.fastcompany.com/magazine/59/mcewen.html">53,000 gold ounces in 1996 to 504,000 gold ounces in 2001</a>, a 851% increase in production. Because extraction cost per ounce
dropped from $360 per ounce to $59 per ounce, an 84% reduction in cost. I
surmise that this reduction in extraction cost is driven by
improvements in extracting more gold from the same amount of ore moved. </li>
</ul>
</div>
<br />
<ul style="text-align: left;">
<li>Accuracy: <i>"We have drilled four of the winner's top five targets and have hit on all four."</i></li>
</ul>
<br />
<br />
<ul style="text-align: left;">
<li>Speed: <i>"McEwen believes that [the challenge] cut two, maybe three years off the company's exploration time."</i></li>
</ul>
<br />
<br />
<ul style="text-align: left;">
<li>1,043,000% ROI: A $575,000 crowdsourcing investment yielded $6 billion in gold.</li>
</ul>
<br />
<div class="MsoNormal">
<br />
<ul style="text-align: left;">
<li>Firm Survival: The open market
price for gold dropped from $400 per ounce in 1996 to $300 per ounce in 2001.
Given that it cost $360 per ounce to extract the gold, Goldcorp would have
shuttered the mine when the prices dropped below extraction costs. Had they NOT
pursued the Goldcorp Challenge the firm would have disappeared.</li>
</ul>
<br />
<div class="MsoNormal">
<br />
<ul style="text-align: left;">
<li>Stock Growth: Now the second largest gold mining company in the world, with 27.5% CAGR in stock price since The 2001 GoldCorp
Challenge</li>
</ul>
</div>
<i><br /></i><br />
<div class="separator" style="clear: both; text-align: center;">
<a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhRrbx1W-aUTNt8r04GqDwD7yfeCu2WGoR2lMxWeZl81pB_e4heSRuFmTDUooJ-5bM8HdeqkyjKPPVGVdW5V-OvxFk9Jv_ebenF95g6awOYfiMGNbTOe2pn_s-pJSMF5IgB3Oxaj-UnoaYU/s1600/Goldcorp+Chart.JPG" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" height="160" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhRrbx1W-aUTNt8r04GqDwD7yfeCu2WGoR2lMxWeZl81pB_e4heSRuFmTDUooJ-5bM8HdeqkyjKPPVGVdW5V-OvxFk9Jv_ebenF95g6awOYfiMGNbTOe2pn_s-pJSMF5IgB3Oxaj-UnoaYU/s400/Goldcorp+Chart.JPG" width="400" /></a></div>
<div style="margin-left: 0.5in; text-indent: 0px;">
<i><br /></i></div>
<i><br /></i><br />
<br />
<ul style="text-align: left;">
<li>Portability of ‘deposit valuation skill’ enables acquisitions:<i> </i>
Companies acquired since the GoldCorp
challenge include Glamis Gold, Barrick Gold Trading, and Gold Eagle Mines.</li>
</ul>
<o:p></o:p><br />
<br />
<b><span style="font-size: large;">Commentary</span></b><br />
<br />
I do think that the Goldcorp challenge is a marvel, and it arose from
questioning such a simple assumption, “Why should we keep the data private?” What perplexes me though is, if Goldcorp ‘struck gold’ on
the five best new sites they dug, why did they ever disclose this? It may be to
comply with disclosure requirements, but the larger prize was testing and
knowing the efficacy of the analytical techniques.<br />
<div class="MsoNormal">
<o:p></o:p></div>
<div class="MsoNormal">
<br /></div>
<div class="MsoNormal">
The winning team was a small quant-geologist company in
Australia. If they were the best of the bunch, why not acquire their company
and use the analytical advantage to buy underperforming mines around the world?<o:p></o:p></div>
<div class="MsoNormal">
Ok, by comparing the results Goldcorp may know that other
teams submitted the same sites so the benefit could be duplicated.<o:p></o:p><br />
<br /></div>
<div class="MsoNormal">
Note, the firm would have closed at least temporarily, had
they not found better yielding deposits to mine (which reduced their cost per
ounce). But who would ever stare at an impending bankruptcy, or cessation of
operations and say, “Let’s get some quant-geologists in here to save us.”<o:p></o:p><br />
<br /></div>
<div class="MsoNormal">
Goldcorp never ran a challenge again, because they already
knew who was the best. I imagine that for every acquisition thereafter they
brought in the same awardees from the original contest. Could it be that Goldcorp’s success is strictly because they
used the ore targeting software when acquiring other firms, and thereby
pocketed a premium on each purchase?</div>
<div class="MsoNormal">
<o:p></o:p></div>
<br />
<h1>
Citations</h1>
<div class="MsoNormal">
<span style="font-size: x-small;">“<a href="http://www.fastcompany.com/magazine/59/mcewen.html">He Struck Gold On the Net (Really)</a>” by Linda Tischler.
FastCompany. May 31, 2002.<o:p></o:p></span></div>
<span style="font-size: x-small;"><br />
</span><br />
<div class="MsoNormal">
<span style="font-size: x-small;"><u><a href="http://www.wikinomics.com/book/">Wikinomics</a></u> by Don Topscott and Anthony Williams. Copyright
2010. Penguin Group Publishing.</span></div>
<br />
<a href="http://www.fractaltechnologies.com/case_study_goldcorp_challenge.25.html">http://www.fractaltechnologies.com/case_study_goldcorp_challenge.25.html</a> </div>
</div>CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.comtag:blogger.com,1999:blog-4323599490342956532.post-75454200372599158532012-01-29T01:22:00.000-05:002012-01-29T01:22:55.237-05:00Do Successful Analytics Companies Require Quant CEOs?<div dir="ltr" style="text-align: left;" trbidi="on">
<div style="text-align: left;">
<h1>
Do Successful Analytics Companies Require Quant CEOs?</h1>
</div>
<div style="text-align: left;">
I should have read it long ago, but I finally sank my teeth into <a href="http://www.tomdavenport.com/books.html#coa">Competing On Analytics</a> over the holidays. I'm just as enamored with the subject matter as <a href="http://www.tomdavenport.com/books.html">the author</a>, but as expected, the book contained the same examples of analytics competition that I've been hearing for the past decade! I was happy to finally get details on Progressive Insurance's approach, but there was one novel idea which grabbed me... that <b>without an interested and analytics passionate CEO, a company will never be an analytic competitor.</b></div>
<div style="text-align: left;">
<br /></div>
<div style="text-align: left;">
<h1>
CEO Impact</h1>
</div>
<div style="text-align: left;">
Davenport gives many examples of analytical companies and emphasizes the importance of the CEO. He even explains how an analytical CEO can turn a company into an 'analytical competitor' in half the time otherwise required. The prototypical example of course, is Gary Loveman. What Davenport fails to mention is that Loveman accomplished this rapid transformation by laying off half the management, but he surpassed all expectations for shareholder value creation.<br />
<br />
Davenport emphasizes the importance of the CEO in the following excerpts,<br />
<ul style="text-align: left;">
<li><i>"A committed, passionate CEO can put the organization on the fast track to analytical competition."</i> (113)
</li>
<li><i>"Even if an organization has some quality data available, it must also have executives who are predisposed to fact-based decision making."</i> (109)
</li>
<li><i>"Leadership and senior executive commitment"</i> is listed as one of the <i>"Key Elements In An Analytical Capability"</i> (111)
</li>
</ul>
</div>
<div style="text-align: left;">
I might restate this idea as follows though:<br />
<br />
<div style="text-align: center;">
<b>Without a quant CEO you can't become an analytical competitor.</b></div>
<br /></div>
<div style="text-align: left;">
<h1>
List of Analytical Competitors</h1>
</div>
<b>Quants</b><br />
<ul style="text-align: left;">
<li>Capital One - <a href="http://en.wikipedia.org/wiki/Richard_Fairbank">Richard Fairbank</a>. B.S. Economics, Stanford. MBA,Stanford.</li>
<li>Netflix - <a href="http://en.wikipedia.org/wiki/Reed_Hastings">Reed Hastings</a>. M.S. Computer Science, Stanford.</li>
<li>Google - <a href="http://en.wikipedia.org/wiki/Sergey_Brin">Sergey Brin</a>. Computer Science graduate student, Stanford.</li>
<li>Google - <a href="http://en.wikipedia.org/wiki/Larry_Page">Larry Page</a>. Computer Science graduate student, Stanford.</li>
<li>D.E. Shaw - <a href="http://en.wikipedia.org/wiki/David_E._Shaw">David Shaw</a>. Ph.D. Computer Science, Stanford.
</li>
<li>Cemex - <a href="http://en.wikipedia.org/wiki/Lorenzo_Zambrano">Lorenzo Zambrano</a>. B.S. Mechanical Engineering, ITESM. MBA, Stanford.
</li>
<li>Facebook - <a href="http://en.wikipedia.org/wiki/Mark_Zuckerberg">Mark Zuckerberg</a>. Computer Science dropout, Harvard.</li>
<li>Microsoft - <a href="http://en.wikipedia.org/wiki/Bill_Gates">Bill Gates</a>. Computer Science dropout, Harvard.
</li>
<li>Oakland A's, San Diego Padres - <a href="http://en.wikipedia.org/wiki/Paul_DePodesta">Paul DePodesta</a>. B.S. Economics, Harvard.</li>
<li>Harrah's - <a href="http://en.wikipedia.org/wiki/Gary_Loveman">Gary Loveman</a>. Ph.D. Economics, MIT.</li>
<li>Rennaissance Tech -<a href="http://en.wikipedia.org/wiki/James_Harris_Simons">James Simons</a>. B.S. MIT, Ph.D. Mathematics, Berkeley.</li>
<li>Amazon - <a href="http://en.wikipedia.org/wiki/Jeff_Bezos">Jeff Bezos</a>. B.S. Electrical Engineering, Princeton.</li>
<li>Schneider International - <a href="http://www.cio.com/article/29861/Christopher_Lofgren_s_Transition_from_CIO_to_CEO">Chris Lofgren</a>. Ph.D. Operations Research, Georgia Tech.</li>
<li>Wal-Mart- <a href="http://en.wikipedia.org/wiki/Mike_Duke">Mike Duke</a>. B.S. Industrial Engineering, Georgia Tech.</li>
<li>SAS - <a href="http://en.wikipedia.org/wiki/James_Goodnight">Jim Goodnight</a>. Ph.D. Statistics, North Carolina State.
</li>
<li>American Airlines (Yield Management) - <a href="http://en.wikipedia.org/wiki/Robert_Crandall">Robert Crandall</a>. MBA Wharton, B.S. U of RI.</li>
<li>Progressive Insurance - <a href="http://people.forbes.com/profile/glenn-m-renwick/33853">Glenn Renwick</a>. M.S.E. University of Florida.</li>
</ul>
<b>Non-Quants (at least I don't think they are)</b><br />
<ul style="text-align: left;">
<li>Capital One - <a href="http://en.wikipedia.org/wiki/Nigel_Morris">Nigel Morris.</a> MBA London Business School.
</li>
<li>Oakland A's - <a href="http://en.wikipedia.org/wiki/Billy_Beane">Billy Beane</a>. Attended UCSD, recruited by Stanford before going pro.
</li>
<li>US GoldCorp Inc. - <a href="http://en.wikipedia.org/wiki/Rob_McEwen">Rob McEwen.</a> MBA Schulich, B.A. Western Ontario.
</li>
<li>E.J. Gallo Winery - <a href="http://en.wikipedia.org/wiki/Joseph_Edward_Gallo">Joe Gallo</a>. Modesto Junior College.
</li>
</ul>
<div style="text-align: left;">
The presence of a quant CEO looks mandatory for the creation of an analytical competitor based on this list, with the rare exceptions of Goldcorp's CEO (who pioneered crowdsourcing), Billy Beane (a pro baseball player), Joe Gallo (the oldest entry on the list by a forty year margin), and two people whose undergraduate studies I cannot determine. I would also point out that Economics degrees at Stanford, Harvard, and MIT are very similar in content to statistics or mathematics degrees at other universities... clearly making these people 'quants'.</div>
<div style="text-align: left;">
<br /></div>
<div style="text-align: left;">
The other item is that many of these quant CEOs, of the most powerful analytical competitors <u>are also founders.</u> Google, Netflix, Facebook, Amazon, RenTech, D.E. Shaw, SAS, and Microsoft. Why? I think for two significant reasons: first, quants are devotees of Data Driven Decision Making, which gives companies they found an operational advantage; second, quants rarely make it to the CEO role because the quant perspective is antethetical to corporate politics. Quants rarely get into executive positions when not founders, but when they do the results are substantial. Think about it... can you picture your company promoting anyone who looks or acts like Mark Zuckerberg, Bill Gates or Jeff Bezos? But these are precisely the people that revolutionize industries.</div>
<div style="text-align: left;">
<br />
<h1>
Commentary</h1>
</div>
<div style="text-align: left;">
I was surprised that AT+T or any of the telecom companies didn't make Davenport's list. I was tempted to add them because data mining texts abound with examples of churn prediction and small business identification from call behavior, but I don't think that any telco has differentiated themselves as a result. Accordingly, it is difficult to single even one of them out as an analytical competitor when the predominant differentiator of performance is iPhone availability. <br />
<br />
Being an analytical competitor requires Data Driven Decision Making. As mentioned in my <a href="http://cavqm.blogspot.com/2011/10/microsoft-exp-and-experimentation-roi.html">Experimentation ROI blog</a>, intuition is wrong 80% of the time so the DDDM is needed before analytics is respected, applied, and reaped.</div>
<div style="text-align: left;">
<br />
<b>Rule of Thumb: To become an analytical competitor, your CEO must have taken multivariable calculus.</b></div>
<div style="text-align: left;">
<div>
<br /></div>
<div>
<br /></div>
<br />
<br /></div>
<div style="text-align: left;">
<span style="font-size: x-small;"><u>Competing On Analytics</u> by Thomas Davenport. Copyright 2007 Harvard Business School Publishing.</span></div>
<div style="text-align: left;">
<br /></div>
<div style="text-align: left;">
</div>
</div>CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.comtag:blogger.com,1999:blog-4323599490342956532.post-73976995168498142732011-12-10T22:53:00.001-05:002012-06-02T09:52:12.023-04:00Normality Assumptions and Power Law Distributions<div dir="ltr" style="text-align: left;" trbidi="on">
<div style="text-align: left;">
I was reading through <u><a href="http://www.gladwell.com/dog/index.html">What the Dog Saw</a></u> by Malcolm Gladwell, and came across the following <a href="http://www.gladwell.com/2006/2006_02_13_a_murray.html">story</a>, originally published by the New Yorker in 2006. The story offers three powerful examples demonstrating a fundamental misconception in problem solving (although, it is not described in such dry terms in the article).</div>
<div style="text-align: left;">
<br /></div>
<div style="text-align: left;">
</div>
<a name='more'></a><br />
<br />
<div style="text-align: left;">
<h1>
Assuming Normality</h1>
</div>
<div style="text-align: left;">
Gladwell points to the human tendency to assume all problems are defined by a normal distribution. We've been taught that heights, IQ, stock market returns, diamond clarity, and SAT scores all obey this mean-centered behavior. <i> </i>Even university students adopt this perspective due to a common misinterpretation of the <a href="http://en.wikipedia.org/wiki/Central_limit_theorem">Central Limit Theorem</a> (CLT), that the distribution of values will approach the mean rather than the average of the values will approach the mean. As a result, <i>"The bell-curve assumption has become so much a part of our mental architecture that we tend to use it to organize experience automatically."</i><br />
<i><br /></i></div>
<div style="text-align: left;">
<h1>
Examples of Power Law Distributions in Everyday Life</h1>
</div>
The article starts with the story of 'Million Dollar Murray'. A lovable, destitute, Nevadan alcoholic who is regularly picked up by Reno police, and frequently by an ambulance for detoxification before he is handed over to the police. One officer has picked up Murray several times per week, for each of the 15 years he's been on the police force. He mentions that Murray ran up a $100,000 hospital bill over the past 6 months at the smaller of Reno's two hospitals, realizing that since taxpayers paid the cost of Murray's urgent care <i>"It cost us one million dollars not to do something about Murray." </i>Furthermore, Murray responded positively to probation and became a model citizen, but quickly regressed when his probation ended. With the hospital bills accumulating, Reno could save money by assigning Murray a chaperone to keep him out of the hospital. This begs the question though, "Can some our greatest social problems be solved with inordinate measures targeted at the few?"<br />
<div>
<i><br /></i></div>
<div>
<h1>
Homelessness and the Power Law Distribution</h1>
</div>
<div>
The next example is as profound as it is politically contentious. Dennis Cullane studied homelessness for his doctoral thesis, where he stayed at a shelter for several weeks. He returned two months later and was shocked to not recognize anyone. This led to his first challenge to conventional wisdom, that <i>"the most common length of homelessness is one day." </i>He later created a homeless database so that he could scientifically study the issue, which led to the following insights.</div>
<div>
<ul style="text-align: left;">
<li>80% of shelter visitors are homeless for one night <i>"the most common length of homelessness is one day"</i></li>
<li>10% are episodic users, primarily young drug users who get off the street for a few weeks in winter</li>
<li>10% are chronic homeless, usually older, and many with mental illnesses who rack up exorbitant hospital bills</li>
</ul>
</div>
<div>
<div style="text-align: center;">
<a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhhd7Zor6lB_rAtAaUSgHvLCEaJctCmlXlhaLUULiEZ95QllY1txWiCFkA9IwY0PkS97yLR-zHTHBHe6kqofkj9lhZSxP3PSfSjVq1Mtcon9vAB7NaRY3oWIYouWb0PJm5kbiwKQBVC_fjs/s1600/Normal+Distibution+vs+Power+Law+Distribution.JPG" imageanchor="1" style="clear: left; display: inline !important; margin-bottom: 1em; margin-right: 1em;"><img border="0" height="180" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhhd7Zor6lB_rAtAaUSgHvLCEaJctCmlXlhaLUULiEZ95QllY1txWiCFkA9IwY0PkS97yLR-zHTHBHe6kqofkj9lhZSxP3PSfSjVq1Mtcon9vAB7NaRY3oWIYouWb0PJm5kbiwKQBVC_fjs/s400/Normal+Distibution+vs+Power+Law+Distribution.JPG" width="400" /></a></div>
<br /></div>
<div>
This has profound implications on solving homelessness because our conventional interpretation of 'homelessness' is really 'chronic homelessness.' Redefining the issue as a problem of the few (Power Law view of homelessness), rather than that all homelessness is identical (Normal distribution view of homelessness) we can target and eliminate the real issue.<br />
<b><br /></b><br />
<h1>
Excessive Force in the LAPD and the Power Law Distribution</h1>
The LAPD has endured numerous accusations of excessive force over the years. The unwritten assumption has been that the majority of officers earned the same number of accusations and that they clustered around an average. In fact though, accusations of excessive force per officer obey a power law distribution:<br />
<ul style="text-align: left;">
<li>6,700 officers had zero accusations of excessive force over a four year period</li>
<li>1,400 officers had one or two accusations over a four year period</li>
<li>183 officers had four or more complaints</li>
<li>44 officers had 6 or more complaints</li>
<li>16 officers had 8 or more complaints</li>
<li>1 officer had 16 complaints</li>
</ul>
Absent this data, <i>"you'd propose solutions that would raise the performance of the middle—like better training or better hiring—when the middle didn't need help." </i>The Christopher commission (responsible for investigating the systematic nature of the excessive force allegations) "<i>gives the strong impression that if you fired those forty-four cops the L.A.P.D. would suddenly become a pretty well-functioning police department."</i><br />
<i><br /></i><br />
Think about the significance of this statement... firing 0.5% of the LAPD would instantly convert it into an exemplary law force whereas training programs would doubtfully produce concrete results. Isn't the immediate, concrete solution a powerful tool for improvement? Doesn't this tool arise directly because we've challenged the normality assumption?<br />
<i><br /></i></div>
<div>
<h1>
Carbon Monoxide Emissions and the Power Law Distribution</h1>
</div>
<div>
Auto emissions are another example of power law distributions that offer a concrete solution to a broad social issue. The example begins with the statement that <i>"A 2004 Subaru in good working order has an exhaust stream that's just .06 per cent carbon monoxide, which is negligible" </i>and the assertion that 90% of vehicles don't need an emissions inspection. <i>"But on almost any highway, for whatever reason—age, ill repair, deliberate tampering by the owner—a small number of cars can have carbon-monoxide levels in excess of ten per cent, which is almost two hundred times higher."</i><i style="font-style: normal;"> </i>Heavily used vehicles such as taxi cabs epitomize the offenders, with one example of a cab that emitted more than its weight in annual carbon monoxide emissions. If<i> "five per cent of the vehicles on the road produce fifty-five per cent of the automobile pollution" </i>then shouldn't we focus more on those five percent, rather than inspect every car owner? Would Greenpeace achieve their goals faster by targeting cab companies or perhaps even by giving replacement cars to the worst offenders?</div>
<div style="text-align: left;">
<br />
<div style="text-align: left;">
<h1>
Moral Hazard and Second Order Effects</h1>
I find it interesting that some of the most seemingly intractable problems facing society could be solved immediately by targeting specific individuals. People expect these problems to be moon shots. Our intuition demands an enormous investment be made in resources and effort to solve these problems, because otherwise we wonder, "Why wasn't it solved before?" </div>
<div style="text-align: left;">
<ul style="text-align: left;">
<li>Could we halve pollution next year if Greenpeace paid to replace the worst emitting vehicles? </li>
<li>Could we solve homelessness by giving the chronic homeless apartments? </li>
<li>Could the LAPD solve excessive force by firing 0.5% of their workforce? </li>
</ul>
Even more shocking, could each of these actions be <i>cheaper</i> than maintaining the status quo? I've already mentioned the example of 'Million Dollar Murray' but what about the societal waste in the vehicle inspection process? Every vehicle is inspected biannually, wasting taxpayers' time, costing money, when more efficient solutions exist? How much is spent by LA on the legal defense of excessive force accusations?<br />
<br />
Despite the benefits of immediate resolution and cost efficiency, they violate American standards of equal treatment and moral ideals of self-sufficiency. How can we give an apartment to a homeless person when there are many people who would like that help? For the sake of brevity I'll forego the moral implications, but instead rely on Gladwell's apt summarization of the problem, <i>"Power-law problems leave us with an unpleasant choice. We can be true to our principles or we can fix the problem. We cannot do both."</i><br />
<div>
<br /></div>
</div>
<div style="text-align: left;">
<h1>
Commentary</h1>
</div>
<div style="text-align: left;">
<i>Understanding:</i> I consider this article profound, and note how important it is to <b>understand</b> a problem before attempting to solve it. Similarly it points to our need to evaluate and recognize our assumptions in decision making. It reminds me of an important quote from <a href="http://en.wikipedia.org/wiki/Eliyahu_M._Goldratt">Eliyahu Goldratt</a>'s <a href="http://en.wikipedia.org/wiki/The_Goal_(novel)">The Goal</a>, that people's arguments are usually correct but their assumptions are wrong. Or to loosely paraphrase Deming's <a href="http://mitpress.mit.edu/catalog/item/default.asp?ttype=2&tid=3261">Out of the Crisis</a>, "Policies often outlive the circumstances they were instituted to address." These all point to the need to re-examine our understanding, and our assumptions to assure their validity.<br />
<br />
<i>Is employee value described by a power law distribution or a normal distribution?</i><br />
I was struck by a recent Zuckerberg <a href="http://www.nytimes.com/2011/05/18/technology/18talent.html?pagewanted=all">quote</a>, <span style="background-color: white; font-family: georgia, 'times new roman', times, serif; font-size: 15px; line-height: 22px;"><i>“Someone who is exceptional in their role is not just a little better than someone who is pretty good,” he said. “They are 100 times better.” </i>Even Paul Allen, the cofounder of Microsoft, wrote in his autobiography, <i>"A great programmer can outproduce an average one by ten to one; with a genius, the ratio might be fifty to one."</i> I would reinterpret this as a 'Power Law view of employee value'. I've argued for some time that employees with scalable skill sets are worth more than those with unscalable skill sets (an idea I picked up from Dr. Lester Thurow in <a href="http://www.amazon.com/Zero-Sum-Society-Distribution-Possibilities-Economic/dp/0465085881">Zero Sum Society</a>), but I've never seen this reflected in salary information. What is the value of someone (Jeff Bezos) who builds a company (Amazon) of 33,700 employees that will soon surpass Wal-mart (572,000 employees) in revenues? Blockbuster laughed when Reed Hastings tried to sell them his patent, and instead created Netflix. What is the value of his insight and that patent now? <u>I would argue that the abnormal returns that entrepreneurs can and have earned over the past three centuries are proof that employee value is a power law distribution.</u> This in turn indicates that most companies seek abnormal returns by paying employees according to a normal distribution, and that </span><span style="background-color: white; font-family: georgia, 'times new roman', times, serif; font-size: 15px; line-height: 22px;">Zuckerberg just wants to arbitrage the market inefficiency.</span><br />
<span style="background-color: white; font-family: georgia, 'times new roman', times, serif; font-size: 15px; line-height: 22px;"><i><br /></i></span><br />
<span style="background-color: white; font-family: georgia, 'times new roman', times, serif; font-size: 15px; line-height: 22px;"><i>Evaluate Individuals Rather Than Populations</i></span><br />
<span style="background-color: white; font-family: georgia, 'times new roman', times, serif; font-size: 15px; line-height: 22px;">Capital One was the first credit card company to use microsegmentation and they were the fastest growing stock on the SP500. Progressive Insurance regularly targets low-risk individuals within unattractively risky populations such as elderly motorcycle owners vs the general motorcycle owner market. If microsegmentation has been fine tuned for customer application, then why have no governments applied it to social problems, or companies applied it to recruiting? It seems like HR would would be an ideal area of application, especially considering it's already been done in professional sports... so what's the hold up?</span><br />
<br /></div>
<div style="text-align: left;">
<br />
<br /></div>
<div style="text-align: left;">
<br /></div>
<div style="text-align: left;">
<span style="font-size: x-small;">"Million Dollar Murray: Why problems like homelessness may be easier to solve than to manage." by Malcolm Gladwell. February 13, 2006. The New Yorker.</span></div>
<div style="text-align: left;">
<br />
<a href="http://orionwell.files.wordpress.com/2007/06/random-vs-power-law-distribution-2.jpg">http://orionwell.files.wordpress.com/2007/06/random-vs-power-law-distribution-2.jpg</a>
<br />
<br />
<a href="http://en.wikipedia.org/wiki/Eliyahu_M._Goldratt">http://en.wikipedia.org/wiki/Eliyahu_M._Goldratt</a>
<br />
<br />
<a href="http://mitpress.mit.edu/catalog/item/default.asp?ttype=2&tid=3261">http://mitpress.mit.edu/catalog/item/default.asp?ttype=2&tid=3261</a>
<br />
<br />
"For Buyers of Web Startups, Quest to Corral Young Talent." By Miguel Helft. New York Times. May 17, 2011.<br />
<br />
<a href="http://www.amazon.com/Zero-Sum-Society-Distribution-Possibilities-Economic/dp/0465085881">http://www.amazon.com/Zero-Sum-Society-Distribution-Possibilities-Economic/dp/0465085881</a>
<br />
<br />
Idea Man: A Memoir by the Cofounder of Microsoft. by Paul Allen. Published by Penguin Group, 2011. New York, New York. </div>
</div>
</div>CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.comtag:blogger.com,1999:blog-4323599490342956532.post-39384910306937788062011-11-30T22:03:00.001-05:002011-12-17T12:57:37.963-05:00Moore's Law and Computer Simulation in Microchip Design<div dir="ltr" style="text-align: left;" trbidi="on">
<div style="text-align: left;">
Many people have heard of <a href="http://en.wikipedia.org/wiki/Moore%27s_law">Moore's Law</a>, that <i>"the number of transistors that can be placed inexpensively on a microchip will double every two years" </i>and assumed that it drove the performance improvements in microchips over the past forty years.</div>
<div class="separator" style="clear: both; text-align: center;">
<a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhrNvxBHBXuCfoax5Uzet3iW2DW3wFmmTBVB4rwFiAqozoCZDcQHv1vSjHnwxXpYxhDMVgbwv26n4T5rG7BEPjdqE0gk_Pok4ah0ZW_zUqcm078GbKBFnwLzR7owt2htK2iSOfNRE1MxXU1/s1600/moore.png" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" height="287" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhrNvxBHBXuCfoax5Uzet3iW2DW3wFmmTBVB4rwFiAqozoCZDcQHv1vSjHnwxXpYxhDMVgbwv26n4T5rG7BEPjdqE0gk_Pok4ah0ZW_zUqcm078GbKBFnwLzR7owt2htK2iSOfNRE1MxXU1/s320/moore.png" width="320" /></a></div>
<div style="text-align: left;">
Only recently did I learn the role that computer simulation and <a href="http://en.wikipedia.org/wiki/Queueing_theory">queueing theory</a> had in the performance improvements. It turns out that microchips were <i>really</i> inefficient at first. For example, </div>
<div style="text-align: left;">
</div>
<ul style="text-align: left;">
<li>The microchip runs faster than the hard drive and non-cache memory can deliver inputs... even today!</li>
<li>Tasks often get stuck behind other tasks, so a 'clog' can drastically impact performance.</li>
<li>Very often the microchip has nothing to do, so the time is wasted.</li>
<li>You can run the exact same routine a dozen times, but the processor didn't realize it could use the answer already calculated.</li>
</ul>
Next I'll describe some of the efforts made to correct these issues.<br />
<div style="text-align: left;">
<br /></div>
<div style="text-align: left;">
<h1><span class="Apple-style-span" style="font-size: large;">Scheduling Improvements</span></h1></div>
<ul style="text-align: left;">
<li><u>Pipelining:</u> CPUs spend much of their time doing arithmetic, so Intel created one pipeline for arithmetic computations and another pipeline for all non-mathematical operations. This proved so successful that most modern microchips have eight different pipelines. (Meyers 141)</li>
<li><u>SRAM/L1 Caching (</u><a href="http://www.tomshardware.com/forum/69310-28-dram-sram">SRAM is 10x-50x faster than regular RAM</a>): The CPU is so fast that regular RAM can't feed it information fast enough, which forces the CPU into wait states. To avoid wait states, Intel put SRAM in Pentium chips to feed the CPU the most commonly calculated pieces of data. <i>"<i>This SRAM would preload as many instructions as possible and would keep copies of already run instructions and data in the hope that the CPU would need to work on them again. </i>The SRAM cache inside the CPU was tiny, only about 16KB, but it improved performance tremendously."</i> (142) </li>
<li><u>L2 Caching: </u>Caching improved performance so much <i>"that motherboard makers began adding a cache directly to the Pentium motherboards."</i> (143) This L2 cache eventually migrated onto the chip from the motherboard. <i>"It's tempting to ask why processor manufacturers didn't just include bigger L1 caches instead of making onboard L1 and L2 caches. The answer is that a very small L1 and a larger L2 are much more efficient than a single fast L1." </i>(143)</li>
<li><a href="http://en.wikipedia.org/wiki/Branch_predictor">Branch prediction:</a> <i>"a process whereby the CPU attempted to anticipate program branches before they got to the CPU"</i> (143) <i>"The [Pentium Pro] improved on the Pentium's branch prediction by adding a far more complex counter that would predict branches with a better than 90-percent success rate." </i>(146)</li>
<li><u>Superscalar Execution: </u><i>"The ability to run more than one instruction in any one clock cycle."</i> (145)</li>
<li><a href="http://en.wikipedia.org/wiki/Out-of-order_execution">Out-of-order processing:</a> <i>"When the [Pentium Pro] was forced into wait states, it took advantage of the wait to look at the code in the pipeline to see if it could run any commands while the wait states were active. If it found commands it could process that were not dependent on the data being fetched from DRAM, it ran these commands out of order, a feature called out-of-order processing." </i></li>
<li><u>Clock Speed Multiplying (up to 30x improvement): </u>The CPU can run between 2x and 30x faster than it can communicate with the rest of the computer.</li>
</ul>
<div style="text-align: left;">
<br />
<div>
<br />
<div>
<b><br /></b></div>
<br />
<h1><span class="Apple-style-span" style="font-size: large;">What is simulation and how can it create advantage?</span></h1></div>
<div>
<br />
<div>
People have used mathematical models of complex systems for decades. The vast majority of these equations rely on assumptions such as <a href="http://en.wikipedia.org/wiki/Memorylessness">memorylessness</a> and the <a href="http://en.wikipedia.org/wiki/Poisson_process">independence of arrival times</a>. These models are fantastic at representing human behavior, financial markets, physics, and natural phenomena but are poor at representing the traffic inside of a computer network. So how do we address this problem? With simulation.<br />
<br />
<div>
</div>
<br />
<br />
<a href="http://en.wikipedia.org/wiki/Computer_simulation">Computer simulation</a> is the use of mathematical models to imitate real world phenomena. It is much cheaper to model a system digitally than to physically perform experiments, and there are many occasions where the mathematical models cannot be solved by hand. The solution is to therefore model the system, run the simulation a million times and examine the distribution of outcomes. Better models will produce a better distribution of outcomes and vice versa.<br />
<br /></div>
</div>
<br />
<h1><span class="Apple-style-span" style="font-size: large;">Commentary</span></h1></div>
<br />
<div>
<u>Non-Intuitive Results: </u>You can convince me that the L1 Cache intuitively makes sense. It's nonintuitive that an L1+L2 is faster than a big L2 though, and I still have difficulty believing that 8 pipelines is faster than 1 for any reason beyond load balancing. Based on these examples, I'd say that simulation is necessary to overcome the limits of intuition when more elegant mathematical solutions also fail.<br />
<br />
<u>Competitive Advantage:</u> Beyond looking at the benefits in hardware terms or product quality, I think we need to ask, "What would have happened if Intel or AMD implemented a queueing improvement that the other didn't realize?" Depending on the time required to retool a factory, such a mistake may have killed them.<br />
<br />
<div>
<br /></div>
<br />
<u>Other Reasons I Like Simulation:</u></div>
<div>
</div>
<br />
<div style="text-align: left;">
<ul style="text-align: left;">
<li>Simulation is a form of experimentation.</li>
<li>Simulation is a form of testing.</li>
<li>Simulation is cheap.</li>
<li>Simulation speeds up product development.</li>
<li>Simulation improves learning.</li>
<li>Simulation is the easiest way to model and understand complex systems.</li>
<li>Simulation is the easiest way to model dependent systems.</li>
</ul>
<div style="text-align: left;">
<br /></div>
<div style="text-align: left;">
<br />
QM-ROI: By my count, queueing theory improvements were responsible for half of the performance improvement in chips over the past 40 years, and queuing theorists are relatively cheap in comparison to a <a href="http://www.azcentral.com/business/articles/2011/02/18/20110218chandler-arizona-announces-new-Intel-facility.html">$5 billion dollar chip plant</a>. Needless to say, anytime you're investing a billion dollars in a physical system that involves scheduling: get expert advice. It could save you millions in factory productivity, hardware performance, logistics expense, etc.<br />
<div>
</div>
</div>
<div style="text-align: left;">
<br />
<br />
<br />
<br /></div>
<div style="text-align: left;">
<span class="Apple-style-span" style="font-size: xx-small;"><br /></span></div>
<div style="text-align: left;">
<span class="Apple-style-span" style="font-size: xx-small;"><u><a href="http://www.amazon.com/Mike-Meyers-Guide-Managing-Troubleshooting/dp/0072231467">Managing and Troubleshooting PCs</a></u> by Mike Meyers. Third Edition. Copyright 2010. Published by McGraw-Hill.</span></div>
<div style="text-align: left;">
<span class="Apple-style-span" style="font-size: xx-small;"><br /></span></div>
<div style="text-align: left;">
<a href="http://en.wikipedia.org/wiki/Moore%27s_law"><span class="Apple-style-span" style="font-size: xx-small;">http://en.wikipedia.org/wiki/Moore%27s_law</span></a></div>
<div style="text-align: left;">
<span class="Apple-style-span" style="font-size: xx-small;"><br /></span></div>
<div style="text-align: left;">
<a href="http://en.wikipedia.org/wiki/Poisson_process"><span class="Apple-style-span" style="font-size: xx-small;">http://en.wikipedia.org/wiki/Poisson_process</span></a></div>
<div style="text-align: left;">
<span class="Apple-style-span" style="font-size: xx-small;"><br /></span></div>
<div style="text-align: left;">
<a href="http://en.wikipedia.org/wiki/Computer_simulation"><span class="Apple-style-span" style="font-size: xx-small;">http://en.wikipedia.org/wiki/Computer_simulation</span></a></div>
<div style="text-align: left;">
<span class="Apple-style-span" style="font-size: xx-small;"><br /></span></div>
<div style="text-align: left;">
<a href="http://en.wikipedia.org/wiki/Branch_predictor"><span class="Apple-style-span" style="font-size: xx-small;">http://en.wikipedia.org/wiki/Branch_predictor</span></a></div>
<div style="text-align: left;">
<span class="Apple-style-span" style="font-size: xx-small;"><br /></span></div>
<div style="text-align: left;">
<a href="http://en.wikipedia.org/wiki/Out-of-order_execution"><span class="Apple-style-span" style="font-size: xx-small;">http://en.wikipedia.org/wiki/Out-of-order_execution</span></a></div>
<div style="text-align: left;">
<span class="Apple-style-span" style="font-size: xx-small;"><br /></span></div>
<div style="text-align: left;">
<a href="http://www.azcentral.com/business/articles/2011/02/18/20110218chandler-arizona-announces-new-Intel-facility.html"><span class="Apple-style-span" style="font-size: xx-small;">http://www.azcentral.com/business/articles/2011/02/18/20110218chandler-arizona-announces-new-Intel-facility.html</span></a></div>
</div>
</div>CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.comtag:blogger.com,1999:blog-4323599490342956532.post-69702279722973632962011-11-14T23:25:00.001-05:002011-12-17T12:58:05.193-05:00The Dowding System and Information Strategy in the Battle of Britain<meta content="The story of Sir Hugh Dowding and how he used a complex system of communications, radar stations, visualization and fighter flight scheduling to foil Nazi invasion plans." />
<div dir="ltr" style="text-align: left;" trbidi="on">
<div style="text-align: left;">
Sir <a href="http://en.wikipedia.org/wiki/Hugh_Dowding,_1st_Baron_Dowding">Hugh Dowding</a>, the Chief of RAF Fighter Command during the 1940 <a href="http://en.wikipedia.org/wiki/Battle_of_Britain">Battle of Britain</a> faced a critical question, "<i>How to best use the 1,000 fighter aircraft under his command against the much larger Luftwaffe?"</i> He was responsible for the British air defense and knew from WWI that it was rare to find the enemy to engage them. Hoping to stumble on the enemy was futile, because even if the enemy aircrafts' location were known they deliberately changed course to disguise their true bombing targets. Alternatively, the Luftwaffe could overwhelm the RAF with countless flight plans. Improvements in aircraft further complicated matters as speeds exceeded 250 mph, increasing the required speed of communications, analysis, and combat. To counteract his numerical disadvantage Dowding turned to a system which integrated cutting edge communications, information gathering, flight scheduling, target prioritization and interception guidance: The Dowding System.</div>
<div style="text-align: left;">
<br /></div>
<div style="text-align: left;">
<h1>The Dowding System</h1></div>
<div style="text-align: left;">
Dowding's brainchild was called, <i><a href="http://www.experiencefestival.com/a/Battle_of_Britain_-_The_Dowding_System/id/615180">"The complex machinery of detection, command and control that ran the battle."</a> </i>Great Britain was on the cutting edge of fighter aircraft design and manufacturing as WWII opened, but armaments, communications and enemy detection were still in their infancy. Dowding oversaw the installation of ground-to-air and air-to-air radios throughout the fighter squadrons, the testing and installation of radar arrays, the creation of a centralized communications system. It is this communications system which I credit with much of the victory.<br />
<br />
The first requisition by Dowding on appointment to the fighter squadrons was to quintuple the number of phone lines running into his headquarters. He also fortified these lines against enemy bombing and ordered the creation of a custom made table. It may sound peculiar, but he was creating the first real-time interactive battle plan. The table (shown below) was carved with the outline of the English coast, and was surrounded by volunteer 'plotters'. The plotters would update the battle plan in real time to reflect friendly and enemy aircraft movements that were constantly phoned into headquarters.<br />
<div class="separator" style="clear: both; text-align: center;">
<a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhNAYNdKcTLLbne2_rRzTJabQULSOsd-NUY9fcqkwb0j-PZgglGTLYpjm01tUEVucgC9fERlUVInn17zq1w1lzti63e29y7VUYblAxKXJEPHDPOSvvCDOfprt4B7JEngpdh97ixMbJGwki-/s1600/Dowding+Table.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" height="275" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhNAYNdKcTLLbne2_rRzTJabQULSOsd-NUY9fcqkwb0j-PZgglGTLYpjm01tUEVucgC9fERlUVInn17zq1w1lzti63e29y7VUYblAxKXJEPHDPOSvvCDOfprt4B7JEngpdh97ixMbJGwki-/s400/Dowding+Table.jpg" width="400" /></a></div>
<br />
<br />
Radar supplemented the phone reports, especially given its indifference to cloud cover and its ability to detect planes still over France and the Netherlands. These reports were supplemented yet again by altitude estimates from human spotters, owing to the newly implemented radar's inability to detect altitude.<br />
<br />
With this information gathered and processed in this fashion, the plotters could readily provide coordinates, altitudes, and interception courses needed by RAF squadrons while Dowding tried to discern Luftwaffe targets from feints.<br />
<br /></div>
<div style="text-align: left;">
<h1>Quantitative Methods: Scheduling, Signal Detection, and Target Prediction</h1></div>
<ul style="text-align: left;">
<li><i>Maximize Fighter Utilization through Scheduling: </i>Launch fighters so as to maximize time in the air, allowing one British aircraft to fight off multiple waves of enemy aircraft with limited fuel supplies. </li>
<li><i>Target Prediction: </i>Predict the most likely target using the course and enemy aircraft type data.</li>
<li><i>Interception Calculations: </i>With 500 mph closing speeds for aircraft, enemy course prediction needed to be real time and very accurate.</li>
<li><i><i>Trigonometry: </i><span class="Apple-style-span" style="font-style: normal;">Dowding created a network of 1,400 trained spotters who used trigonometry to calculate the altitude of enemy aircraft.</span></i></li>
<li><i>Signal Detection: </i>Identified aircraft on radar readouts, distinguishing friend from foe and determining the number of aircraft</li>
<li><i>Data Visualization: </i>All radar data, spotter observations, aircraft movements and engagements were plotted in real time (picture below).</li>
</ul>
<div style="text-align: left;">
</div>
<b><br /></b><br />
<div style="text-align: left;">
<h1>Quantification of Benefits</h1></div>
<div style="text-align: left;">
</div>
<ul style="text-align: left;">
<li><a href="http://margetfreebush.tripod.com/id5.html">"At times, the Dowding System achieved interception rates exceeding 80%."</a> </li>
<li>The invasion of Great Britain (<a href="http://en.wikipedia.org/wiki/Operation_Sea_Lion">Operation Sea Lion</a>) was postponed indefinitely.</li>
<li>The RAF were outnumbered two to one, but they fought off the Luftwaffe. It could therefore be claimed that the Dowding System was a force multiplier that at minimum doubled their strength.</li>
<li><i>Fuel economy:</i> This may seem trivial, but the RAF fighters used 100 octane fuel imported from the United States. It enhanced the performance of the aircraft but it also made fighter fuel a scarce resource to be optimized, so it wasn't feasible to have aircraft on constant patrol. The Dowding System was fuel optimal.</li>
</ul>
<div style="text-align: left;">
<b><br /></b><br />
<h1>Commentary</h1></div>
<div style="text-align: left;">
I picked up this book to read over the Thanksgiving holiday without realizing the strategic lessons it would offer. It immediately reminded me of the quote, <i>"Wars are won by logistics, communications, and battles... in that order"</i> with particular emphasis on the importance of communications. One of Dowding's first acts upon appointment to lead fighter command was the centralization of communications, the quintupling of the phone lines to that centralized location, reinforcement of those phone lines with concrete to protect them from bombing, and the plotting table. This may seem excessive, but as mentioned in my Cemex blog <i>'data is the foundation of CAvQM'</i> and communications are the conduit of data delivery.<br />
<br />
I was also impressed by the primacy of data collection in the Dowding System. The RAF also needed fighters that could rival the Luftwaffe, but without a very sophisticated data collection and filtering mechanism they could not have used their RAF fighters effectively. Furthermore, the numerical superiority of the Luftwaffe forced the British to accurately distinguish the important sorties from the diversions.<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<a href="http://www.amazon.com/Wings-Like-Eagles-History-Britain/dp/0061125350">With Wings Like Eagles: The Battle of Britain.</a> By Michael Korda. Copyright 2009. Harper Publishing.<br />
<br />
<a href="http://margetfreebush.tripod.com/id5.html">http://margetfreebush.tripod.com/id5.html</a><br />
<br />
http://en.wikipedia.org/wiki/Battle_of_Britain<br />
<br />
http://en.wikipedia.org/wiki/Hugh_Dowding,_1st_Baron_Dowding<br />
<br />
http://www.experiencefestival.com/a/Battle_of_Britain_-_The_Dowding_System/id/615180<br />
<br /></div>
</div>CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.comtag:blogger.com,1999:blog-4323599490342956532.post-82715588894549301682011-10-18T01:12:00.000-04:002011-12-17T13:04:38.782-05:00Microsoft ExP and Experimentation ROI<div dir="ltr" style="text-align: left;" trbidi="on"><br />
<div class="MsoNormal">
<b>
<h1>
<span class="Apple-style-span" style="font-size: large;">Why do you need experimentation? Can’t we just rely on our best judgment?</span>
<o:p></o:p>
</b>
</h1>
</div>
<div class="MsoNormal">Just in case <a href="http://cavqm.blogspot.com/2011/10/ab-testing-and-experimentation-as.html">my prior blog</a> didn't convince you, the answer is no, you can’t rely on your best judgment because <u>our intuition is wrong 80% of the time.</u> This Microsoft <a href="http://exp-platform.com/Documents/ExP_DMCaseStudies.pdf">whitepaper</a> offers the following proof:<o:p></o:p></div><div class="MsoNormal"></div><ul style="text-align: left;"><li><i><span class="Apple-style-span" style="font-style: normal;"><i>“<a href="http://www.qualproinc.com/">QualPro</a>, a consulting company specializing in [experimentation], tested 150,000 business improvement ideas over 22 years and reported that 75 percent of important business decisions and business improvement ideas either have no impact on performance or actually hurt performance.”</i></span></i></li>
<li><i>"Avinash Kaushik, author of <u><a href="http://www.webanalyticshour.com/">Web Analytics: An Hour A Day</a></u>, wrote that ‘80% of the time you/we are wrong about what a customer wants.’”</i></li>
<li><i>“In <u><a href="http://www.mikemoran.com/diwq/">Do It Wrong Quickly</a></u>, the author writes that Netflix considers 90% of what they try to be wrong.”</i></li>
<li><i>“Regis Hadaris from <a href="http://www.quickenloans.com/">Quicken Loans</a> wrote that ‘I’ve been doing this for five years, and I can only ‘guess’ the outcome of a test about 33% of the time!’”</i></li>
<li><i>“In an old classic, <u><a href="http://en.wikipedia.org/wiki/Scientific_Advertising">Scientific Advertising</a></u>, the author writes that ‘[In selling goods by mail] false theories melt away like snowflakes in the sun… One quickly loses his conceit by learning how often his judgment errs – often nine times in ten.’”</i></li>
<li><i>“we can report that Microsoft is no different. Evaluating well-designed and executed experiments that were designed to improve a key metric, only about one-third were successful at improving the key metric!”</i></li>
</ul><br />
<div class="MsoNormal">
<b>
<h1>
<span class="Apple-style-span" style="font-size: large;">Creating A Competitive Advantage</span>
</h1>
</b>
</div>
<div class="MsoNormal">If you’ve warmed to the idea that 80% of business decisions actually hurt performance, the next logical question is ‘How can I avoid this behavior, and maybe create an advantage?’ After all, you could improve decision-making quality 5-fold, permit reallocation of resources to more profitable endeavors, and remove socio-political factors from the process. But if my word doesn’t carry enough weight I’ll defer to the following evangelists of CAvQM:<o:p></o:p></div><div class="MsoNormal"></div><ul style="text-align: left;"><li><i>“Being able to figure out quickly what works and what doesn’t can mean the difference between survival and extinction.”</i> – <a href="http://en.wikipedia.org/wiki/Hal_Varian">Hal Varian</a>, Google Chief Economist</li>
<li><i>“There were three ways to get fired at Harrah’s: steal, harass women, or institute a program or policy without first running an experiment.”</i> – <a href="http://en.wikipedia.org/wiki/Gary_Loveman">Gary Loveman</a>, quoted in <u><a href="http://www.amazon.com/Facts-Dangerous-Half-Truths-Total-Nonsense/dp/1591398622">Hard Facts</a></u>.</li>
</ul><br />
<div class="MsoNormal">Perhaps the most surprising fact is that <u>applying experimental methods produces some of the largest ROI’s I’ve ever seen in the business world</u>.<o:p></o:p></div><div class="MsoNormal"><br />
</div>
<div class="MsoNormal">
<b>
<h1>
<span class="Apple-style-span" style="font-size: large;">Quantification of Benefits – Experimentation ROI</span>
</h1>
<o:p></o:p>
</b>
</div>
<div class="MsoNormal">Did I mention that experimentation is also lucrative? The process is inherently self-funding because bad decisions are expensive. <u>How much would you pay to avoid bad decisions?</u> Fortunately, Microsoft provided a number of project ROI in their white paper so we have reliable numbers.</div><div class="MsoNormal"></div><ul style="text-align: left;"><li><i>“In one case we ran an experiment for a site where the management was reluctant to run the test because they considered it a no-brainer’ that the Treatment would win. The Treatment had some unexpected and subtle negative aspects that would not have been detected had we not run the experiment. If the Treatment had been launched we estimate the annual loss to the site would have been in the millions of dollars. It only cost them $6,000 to run the experiment.”</i></li>
<ul><li><b>Experimentation ROI greater than 33,000%.</b></li>
</ul><li>Amazon Behavior Based Search:<i> “In [another paper] we described the development of Behavior-Based Search at Amazon, a highly controversial idea. Early experiments by an intern showed the surprisingly strong value of the feature, which ultimately helped improve Amazon’s revenue by 3%, translating into hundreds of millions of dollars in incremental sales. <a href="http://glinden.blogspot.com/">Greg Linden</a></i><i> at Amazon created a prototype to show personalized recommendations based on items in the shopping cart. Linden notes that ‘a marketing senior vice-president was dead set against it.’ Claiming it will distract people from checking out. Greg was ‘forbidden to work on this any further.’ Nonetheless, Greg ran a controlled experiment and the rest is history: the feature was highly beneficial. Multiple sites have copied cart recommendations.”</i></li>
<ul><li><b>Experimentation ROI greater than 1,000,000%.</b></li>
</ul><li>Microsoft tested the affect on user experience from adding 3 ads to the MSN home page, which would generate $1 million in advertising revenue annually. Unfortunately, experiments showed that the CTR (click-through rate) would decline by 0.35% if the ads were incorporated, <i>“the estimated loss, had this feature been deployed, was millions of dollars per year."</i></li>
<ul><li><b>Experimentation ROI greater than 50,000%.</b></li>
</ul></ul><div class="MsoListParagraphCxSpLast" style="margin-left: 1.0in; mso-add-space: auto; mso-list: l1 level2 lfo1; text-indent: -.25in;"><o:p></o:p></div><div class="MsoNormal"><b><br />
</b>
</div>
<h1>
<div class="MsoNormal"><b><span class="Apple-style-span" style="font-size: large;">The Mechanics of Building A Competitive Advantage Around Experimentation</span></h1>
<o:p>
</o:p>
</b>
</div>
<div class="MsoNormal"><b>
<br />
</b>
</div>
<div class="MsoNormal"><b>Hiring:</b> Companies that build a competitive advantage around experimentation start by hiring developers and statisticians to provide the subject matter expertise, social credit, and coding ability. Early efforts focus on proof of concept experiments to build momentum, while socializing the benefits, and creating a data driven culture while constructing an experimentation platform. <o:p>
</o:p>
</div>
<div class="MsoNormal">
<b>
<br />
</b>
</div>
<div class="MsoNormal">
<b>Prove The Shortcomings of Intuition: </b>People will not abandon intuitive decision making easily. They need to be shown the benefits of experimentation in contests. Offer $1,000 to any employee who can correctly pick the winner from 6 of 10 experiments (when Microsoft tried this, no one came close). <b>
<o:p></o:p></b>
</div>
<div class="MsoNormal">
<b>
<br />
</b></div>
<div class="MsoNormal"><b>Why invest in an experimentation platform?</b> <i>“You have to kiss a lot of frogs to find a prince. So how </i><i>can you find your prince faster? By finding more frogs and kissing them faster and faster.”</i> <span> </span><u>A robust experimentation platform is needed to </u><i><u>“increase the rate of experimentation and lower the cost to run experiments.</u>”</i><i> </i>Also remember that 80% of the time, executives mistake princes for frogs so many decisions must be tested to avoid marrying frogs.</div>
<div class="MsoNormal"><o:p></o:p></div><div class="MsoNormal"><b><br />
</b></div>
<div class="MsoNormal"><b>Test often, test early.</b> Catching a poor decision, design, or idea early saves development expenses, organizational resources, brand value and customer goodwill. <i>"It is well known that finding an error in requirements is 10 to 100 times cheaper than changing features in a finished product."</i>
<span class="Apple-style-span" style="font-family: Calibri, sans-serif;"><span class="Apple-style-span" style="font-size: 15px; line-height: 17px;"><br />
</span>
</span></div>
<div class="MsoNormal"><br />
</div>
<div class="MsoNormal"><b><span class="Apple-style-span" style="font-size: large;">Commentary</span><o:p></o:p></b></div><div class="MsoNormal">Just to reiterate the most important point, <b>"<u>applying experimental methods produces some of the largest ROI’s I’ve ever seen</u>." </b>Heck, for a 1,000,000% ROI I'll ignore the competitive advantage and just do it for the return!</div><div class="MsoNormal"><br />
</div><div class="MsoNormal"><br />
</div><div class="MsoNormal"><br />
</div><div class="MsoNormal"><br />
</div><div class="MsoNormal">"Online Experimentation at Microsoft." By Ronny Kohavi, Thomas Crook, and Roger Longbotham. June 28th, 2009.</div></div>CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.comtag:blogger.com,1999:blog-4323599490342956532.post-32604959448996141962011-10-09T15:56:00.000-04:002011-12-17T13:10:54.925-05:00A/B Testing and Experimentation as a Competitive Advantage<div dir="ltr" style="text-align: left;" trbidi="on"><span class="Apple-style-span" style="font-family: inherit;"><span style="color: black;">I've touched on the idea of experimentation as a competitive advantage in my <a href="http://cavqm.blogspot.com/2010/08/harrahs-loyalty-system.html">Harrah's</a></span><span style="color: black;">, <a href="http://cavqm.blogspot.com/2011/03/capital-one-and-microsegmentation-your.html">CapOne</a></span><span style="color: black;">, and <a href="http://cavqm.blogspot.com/2011/01/taylorism-applied-to-metals-engineering.html">Taylorism</a> </span><span style="color: black;">blogs, although I've avoided this topic so far because it is <b>overwhelming</b>. Experimentation is central to the scientific method, is a whole discipline within statistics (</span><a href="http://en.wikipedia.org/wiki/Design_of_experiments">
<span style="color: blue;">Design Of Experiments</span></a>
<span style="color: black;">)</span><span style="color: black;">, and is rapidly shaping the course of digital marketing. In one enlightening demonstration of its importance, Jeff Bezos fired a web design firm for changing the website without running an experiment first. The best indicator of experimentation's growing importance though, is the emphasis on it in the </span><a href="http://www.mckinsey.com/mgi/publications/big_data/"><span style="color: blue;">McKinsey Big Data white paper</span></a><span style="color: black;">.</span></span><br />
<span class="Apple-style-span" style="font-family: inherit;"><br />
</span><br />
<div class="MsoNormal" style="line-height: normal; margin: 0in 0in 0pt;">
<span style="color: black; font-family: inherit;">Why all the fuss though over experimentation though? Because <u>experimentation <i style="mso-bidi-font-style: normal;">is</i> data driven decision making</u><u>.</u> If your decision making process doesn’t utilize statistically valid experimention then <i style="mso-bidi-font-style: normal;">you are a dinosaur</i>.<o:p></o:p></span></div><br />
<div class="MsoNormal" style="line-height: normal; margin: 0in 0in 0pt;"><span class="Apple-style-span" style="font-family: inherit;">
<b style="mso-bidi-font-weight: normal;">
<span style="color: black; font-size: 14pt;">
<h1>
<a href="http://en.wikipedia.org/wiki/A/B_testing">What is A/B Testing?</a>
</h1>
</span>
</b>
<b style="mso-bidi-font-weight: normal;"><span style="color: black;"> <o:p></o:p></span></b></span></div><span class="Apple-style-span" style="font-family: inherit;">A/B testing is the first step in creating an experimentation-based competitive advantage, but first we’ll start with an example.</span>
<br />
<span class="Apple-style-span" style="font-family: inherit;"><br />
</span><br />
<div class="MsoNormal" style="line-height: normal; margin: 0in 0in 0pt;"><span style="color: black; font-family: inherit; font-size: 10pt;"><i>A direct marketer sends out 100 letters to potential customers. He randomly assigns 50 customers to Group A (the control) and 50 customers to group B (the test group). The ‘A’ letters are sent using the normal stationery while the ‘B’ letters are sent using ornate, colored and watermarked stationery. The customers that respond to the letters are tracked when they place an order, and the marketer tallies the results. If Group B had a significantly higher response rate than Group A, they will adopt this stationery as standard.</i><o:p></o:p></span></div><span class="Apple-style-span" style="font-family: inherit;"><br />
</span><br />
<div class="MsoNormal" style="line-height: normal; margin: 0in 0in 0pt;">
<span class="Apple-style-span" style="font-family: inherit;">This technique enables objective comparison of alternatives and makes iterative, incremental improvements possible. Although the method was pioneered in direct mail marketing, it has become a standard tool in digital marketing. One extreme example involved <a href="http://www.nytimes.com/2009/03/01/business/01marissa.html?pagewanted=1">Google deciding between two shades of blue</a> for a webpage. True to their reputation, they A/B tested 41 different shades of blue to determine which inspired the most clicks. </span></div><span class="Apple-style-span" style="font-family: inherit;"><br />
</span><br />
<div class="MsoNormal" style="line-height: normal; margin: 0in 0in 0pt;">
<span class="Apple-style-span" style="font-family: inherit;">
<span style="color: black;">Google and Amazon aren’t the only big names using this technique though. <a href="http://www.hackingnetflix.com/2010/11/netflixs-neil-hunt-says-that-netflix-ab-tests-everything.html">Netflix</a></span><span style="color: black;">, Zynga, <a href="http://thinkvitamin.com/design/how-eharmony-kills-the-romance-with-ab-testing/">eHarmony</a></span> <span style="color: black;">and Microsoft add to the list... in addition to every internet advertiser using the <a href="http://www.google.com/websiteoptimizer">Google Website Optimizer</a>. </span></span></div><span class="Apple-style-span" style="font-family: inherit;"><br />
</span><br />
<div class="MsoNormal" style="line-height: normal; margin: 0in 0in 0pt;">
<span class="Apple-style-span" style="font-family: inherit;">
<b>
<h1>
<span style="color: black; font-size: 14pt;">Moving Beyond A/B Testing</span>
<h1>
</b>
</span>
</div>
<span class="Apple-style-span" style="font-family: inherit;">A/B Testing is ideally suited to comparing two alternatives. It is the first tool in the experimenter’s toolkit, but it is easily overwhelmed for non-incremental comparisons. For example, if you want to know the optimal conditions for car engine performance, you would need to vary temperature, humidity, air pressure, fuel octane and compare the results. There are thousands of such combinations though, and we need more powerful tools to answer such questions. I will save discussion of more sophisticated methods until a later blog though.</span>
<br />
<span class="Apple-style-span" style="font-family: inherit;">
<br />
</span>
<br />
<h1>
<div class="MsoNormal" style="line-height: normal; margin: 0in 0in 0pt;">
<span style="color: black; font-family: inherit; font-size: 14pt; font-weight: bold;">Quantification of Benefits</span>
</h1>
<br />
<span class="Apple-style-span" style="font-family: inherit;">A recent New York Times article, <a href="http://www.nytimes.com/2011/04/24/business/24unboxed.html">“When There’s No Such Thing As Too Much Information”</a> showed that companies using data driven decision-making (DDDM) were 6% more productive than could be explained by other factors. I would point to experimentation as the differentiator by using the following success stories from Google Website Optimizer:</span></div><br />
<ul style="text-align: left;"><li><span lang="EN" style="color: black; font-size: 10pt;">Mattress Liquidators increased online leads 5,000%</span>
</li>
<li><span lang="EN" style="font-size: 10pt;">Doba.com increased conversions 70% and sign-ups 50%</span></li>
<li><span lang="EN" style="font-size: 10pt;">Jigsaw Health increased conversions 60%</span></li>
<li><span lang="EN" style="font-size: 10pt;">Tourism BC boosted conversion rates by 7 percent over previous campaigns</span></li>
<li><span lang="EN" style="color: black; font-size: 10pt;">BC Finance saw an increase in conversion rates exceeding 15 points</span></li>
<li><span lang="EN" style="color: black; font-size: 10pt;">Moishe's Moving Systems increases its lead collection rate by 50%</span></li>
<li><span lang="EN" style="color: black; font-size: 10pt;">Dr. Gary Berger improved conversions 225% with Google Website Optimizer</span></li>
<li><span lang="EN" style="color: black; font-size: 10pt;">Calyx Flowers used Website Optimizer to drive a 14% increase in the number of customers adding items to cart</span></li>
</ul><br />
<span class="Apple-style-span" style="font-family: inherit;">For additional clarification, please note that the Google Website Optimizer is free and turn-key, so every performance improvement went straight to the bottom line.</span><br />
<b><span style="color: black; font-family: inherit; font-size: 14pt;"><br />
</span></b><br />
<b>
<h1>
<span style="color: black; font-family: inherit; font-size: 14pt;">Commentary</span>
</h1>
</b>
<br />
<span class="Apple-style-span" style="font-family: inherit;"><br />
</span><br />
<span class="Apple-style-span" style="font-family: inherit;"><span style="color: black; font-weight: bold;">Data Driven Decision Making Requires a Culture of <u>Discipline</u>: </span><span style="color: black;">People don’t want to abandon their intuition when making decisions, so the corporate culture needs to encourage them to use DDDM.<b> </b>Furthermore, accurately quantifying the effect of marketing campaigns requires that you leave many customers out as a control group. Foregoing these incremental revenues for the sake of quantification accuracy will always be a difficult sell and will require cultural support.</span></span><br />
<span class="Apple-style-span" style="font-family: inherit;"><br />
</span><br />
<span class="Apple-style-span" style="font-family: inherit;"><b><span style="color: black;">Led Astray By Economics Classes:</span></b><span style="color: black;"> So many business people learned economics that we’ve lost touch with the scientific method. This is because <u>much of economics is unscientific</u>. For example, the axiom of the rational consumer was first contradicted by <a href="http://en.wikipedia.org/wiki/Prospect_theory">Kahnemann and Tversky</a>, is now regularly refuted in <a href="http://en.wikipedia.org/wiki/Behavioral_economics">behavioral economics</a>, the axiom of ‘normally distributed financial market returns’ was refuted by <a href="http://en.wikipedia.org/wiki/Eugene_Fama">Eugene Fama</a> and <a href="http://en.wikipedia.org/wiki/Nassim_Nicholas_Taleb">Nassim Taleb</a></span><span style="color: black;">, while <a href="http://www.blogger.com/[http://en.wikipedia.org/wiki/John_Forbes_Nash,_Jr.">John Nash</a> proved that pursuit of self-interest does not lead to market-optimal outcomes. And yet none of these blatant contradictions prevents laissez-faire economics or ‘rational consumerism’ from being taught in economics courses.</span></span><br />
<span class="Apple-style-span" style="font-family: inherit;"><br />
</span><br />
<span class="Apple-style-span" style="font-family: inherit;">There is no other ‘quantitative’ discipline where the axioms have proven so flawed. This is largely because nothing in economic theory is tested. ‘Laws’ are ascribed to historical patterns which do not hold in different time periods or countries. For example, the <a href="http://en.wikipedia.org/wiki/Phillips_curve">Phillips curve</a> is taught in every macroeconomics class, but data has never supported the existence of this law outside of the original data set that was used.</span><br />
<span class="Apple-style-span" style="font-family: inherit;"><br />
</span><br />
<span class="Apple-style-span" style="font-family: inherit;">Is it any wonder then, that after this training, most business people do not experiment but seek interpretations of historical data to support their conclusions?</span><br />
<span class="Apple-style-span" style="font-family: inherit;"><br />
</span><br />
<span class="Apple-style-span" style="font-family: inherit;"><b><span style="color: black;">Local Optima vs. Global Optima: </span></b>A/B testing is ‘iterative optimization’. It allows you to improve a tiny bit every day, but eventually your results reach an optimum and it stops improving. The question then is, “How do we find a new starting point?” The lucrative way to answer this question is with <u>vision</u>. Amazon Silk achieved a 200x improvement in browser speed by re-thinking the browser. Apple took over the smartphone market by re-thinking the phone rather than by creating a slightly improved BlackBerry. These radical improvements do not result from iterative optimization, but from insight and new technology.</span><br />
<span class="Apple-style-span" style="font-family: inherit;"><br />
</span><br />
<div class="MsoNormal" style="margin: 0in 0in 10pt;"><span class="Apple-style-span" style="font-family: inherit;"><b style="mso-bidi-font-weight: normal;"><span style="font-size: 12pt; line-height: 115%;"><a href="http://www.joshhannah.com/2010/07/ab-testing-can-you-iterate-your-way-to-great-products/">‘Splittesting is no substitute for product vision.’</a></span></b><b style="mso-bidi-font-weight: normal;"> </b></span></div><span class="Apple-style-span" style="font-family: inherit;"><br />
</span><br />
<div class="MsoNormal" style="margin: 0in 0in 10pt;"><span class="Apple-style-span" style="font-family: inherit;"><br />
</span></div><span class="Apple-style-span" style="font-size: xx-small;"><span class="Apple-style-span" style="font-family: inherit;">“Putting a Bolder Face on Google.” By Laura M. Holson. The New York Times. February 29, 2009.</span></span><br />
<span class="Apple-style-span" style="font-family: inherit;"><br />
</span><br />
<span class="Apple-style-span" style="font-family: inherit;"><span style="font-size: xx-small;"> </span><span style="font-size: xx-small;">“When There’s No Such Thing as Too Much Information.” By Steve Lohr. The New York Times. April 23, 2011. </span></span><br />
<span class="Apple-style-span" style="font-family: inherit;"><br />
</span><br />
<span class="Apple-style-span" style="font-family: inherit;"><span style="font-size: xx-small;"> </span><a href="http://www.hackingnetflix.com/2010/11/netflixs-neil-hunt-says-that-netflix-ab-tests-everything.html"><span style="color: blue; font-size: xx-small;">http://www.hackingnetflix.com/2010/11/netflixs-neil-hunt-says-that-netflix-ab-tests-everything.html</span></a></span><br />
<span class="Apple-style-span" style="font-family: inherit;"><br />
</span><br />
<span style="font-family: inherit; font-size: xx-small;"><a href="http://thinkvitamin.com/design/how-eharmony-kills-the-romance-with-ab-testing/"> </a><span style="color: blue;"><a href="http://thinkvitamin.com/design/how-eharmony-kills-the-romance-with-ab-testing/">http://thinkvitamin.com/design/how-eharmony-kills-the-romance-with-ab-testing/</a></span><o:p></o:p></span><br />
<span style="font-family: inherit; font-size: xx-small;"><br />
</span><br />
<span style="font-family: inherit; font-size: xx-small;"> </span><br />
<div class="MsoNormal" style="margin: 0in 0in 10pt;"><span class="Apple-style-span" style="font-family: inherit;"><a href="http://www.joshhannah.com/2010/07/ab-testing-can-you-iterate-your-way-to-great-products/"><span style="color: blue; font-size: xx-small;">http://www.joshhannah.com/2010/07/ab-testing-can-you-iterate-your-way-to-great-products/</span></a><o:p></o:p></span></div><span class="Apple-style-span" style="font-family: inherit;"><span style="font-size: xx-small;"> </span><span style="color: black;"><span style="font-size: xx-small;"><a href="http://en.wikipedia.org/wiki/A/B_testing">http://en.wikipedia.org/wiki/A/B_testing</a></span></span></span><br />
<span class="Apple-style-span" style="font-family: inherit;"><br />
</span><br />
<span style="font-size: xx-small;"><span class="Apple-style-span" style="font-family: inherit;"> <a href="http://en.wikipedia.org/wiki/Design_of_experiments"><span style="color: blue;">http://en.wikipedia.org/wiki/Design_of_experiments</span></a></span><span class="Apple-style-span" style="font-family: 'Times New Roman', serif;"><o:p></o:p></span></span><br />
<span style="font-size: xx-small;"> </span></div>CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.comtag:blogger.com,1999:blog-4323599490342956532.post-89097630100226671352011-09-19T22:49:00.000-04:002011-12-17T13:18:32.467-05:00McKinsey Consulting: “Big Data Is The Next Frontier”<div dir="ltr" style="text-align: left;" trbidi="on">
<b style="mso-bidi-font-weight: normal;">
<h1<
<span style="font-size: 14pt; line-height: 115%;"><span style="font-family: Calibri;">What Is Big Data?<o:p>
</h1>
</o:p></span>
</span></b>
<br />
<span style="font-family: Calibri;"><a href="http://en.wikipedia.org/wiki/Big_data">BigData</a> in its literal sense, refers to data whose scale requires radically new tools, storage, software, and approaches to use it fully. <span style="font-size: xx-small;">(1)</span> Big Data now refers to the underlying data sets as well as the analytics driven by it, such as neural networks, artificial intelligence, rules-based systems, and more. <span style="font-size: xx-small;">(99)</span> For purposes of this blog, <u>Big Data is QM</u>, and we will examine it through the lens of McKinsey consulting's 150 page white paper <a href="http://www.mckinsey.com/mgi/publications/big_data/pdfs/MGI_big_data_full_report.pdf">Big data: The next frontier for innovation, competition, and productivity</a>.</span><br />
<br />
<span style="font-family: Calibri;">BIG DATA CONFERS COMPETITIVE ADVANTAGE: McKinsey's <a href="http://www.mckinsey.com/mgi/publications/big_data/pdfs/MGI_big_data_full_report.pdf">paper</a>
</span><span style="font-family: Calibri;">clearly points to the strategic implications of big data. For example, <i style="mso-bidi-font-style: normal;">“The use of big data is becoming a key way for leading companies to outperform their peers.”</i> <span style="font-size: xx-small;">(6)</span> Or similarly, <i style="mso-bidi-font-style: normal;">“…the impact of developing a superior capacity to take advantage of big data will confer enhanced competitive advantage over the long term and is therefore well worth the investment to create this capability.”</i> <span style="font-size: xx-small;">(6)</span><span style="mso-spacerun: yes;"> </span>To bring this point back to the present though, McKinsey cites Tesco, Amazon, Wal-mart, Harrah’s, Capital One, and Progressive Insurance as innovators in this area, partly attributing their success to it. <span style="font-size: xx-small;">(23)<o:p></o:p>
</span>
</span><br />
<br />
<span style="font-family: Calibri;">DATA IS A FACTOR OF PRODUCTION: Traditional economics defines the factors of production as: labor and capital. McKinsey makes an interesting point by proposing ‘data’ as a new factor of production. <span style="font-size: xx-small;">(4)<o:p></o:p></span></span><br />
<br />
<span style="font-family: Calibri;">BIG DATA WILL TRANSFORM ECONOMIES ON PAR WITH THE IT REVOLUTION: <i style="mso-bidi-font-style: normal;">“The same preconditions that explain the impact of IT in enabling historical productivity growth currently exist for big data.”</i> <span style="font-size: xx-small;">(24)<b style="mso-bidi-font-weight: normal;"><o:p></o:p></b></span></span><br />
<br />
<br />
<b style="mso-bidi-font-weight: normal;"><span style="font-size: 14pt; line-height: 115%;">
<h1>
<span style="font-family: Calibri;">Big Data As Strategy</span>
</h1>
</span></b><br />
<span style="font-family: Calibri;">McKinsey has already cited big data (aka QM) as transformational, a digital-age factor of production, and a source of competitive advantage. McKinsey also points to big data as a strategy though, <em>“established competitors and new entrants alike will leverage data-driven strategies to innovate, compete, and capture value.” <span style="font-size: xx-small;">(6) </span></em></span><span style="font-size: 14pt; line-height: 115%;"><span style="font-family: Calibri;"><span style="font-family: Times New Roman; font-size: small;"><span style="mso-spacerun: yes;">Perhaps more interesting though, is the idea that western economies will soon witness an explosion of innovations surrounding big data, </span><em>“…our research suggests that the scale and scope of changes that big data are bringing about are at an inflection point, set to expand greatly, as a series of technology trends accelerate and converge.” <span style="font-size: xx-small;">(2)</span> </em>What happens to industries where data-driven companies see productivity and profitability double? Will the laggards have time to hire quants and catch up? I doubt it.</span></span></span><br />
<br />
<b style="mso-bidi-font-weight: normal;">
<h1>
<span style="font-family: Calibri;">The Coming Shortage of Big Data Talent </span>
</h1>
</b><br />
<span style="font-family: Calibri;">Another profound data point from the article is the coming shortage of big data talent. <i style="mso-bidi-font-style: normal;">“The United States alone faces a shortage of 140,000 to 190,000 people with deep analytical skills as well as 1.5 million managers and analysts to analyze big data and make decisions based on their findings.”</i> <span style="font-size: xx-small;">(3)</span> This assertion fits with a recent WSJ article <a href="http://online.wsj.com/article/SB10001424053111903885604576486330882679982.html">"Business Schools Plan Leap into Data"</a> which cites the importance that MBA programs are placing on the analytical aptitude of their graduates.</span><br />
<br />
<div class="separator" style="clear: both; text-align: center;">
<a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgsKGODCbgPRmCskubXbwyMktzhnC9rPVQSxk6uuJ-bq4k-uwSk0GRiEwwzZHIlb9Rx0iwRVoH4rlJLviUlpzD9SB3X3am4Yfxl3mt7t87_JGgc5ayKcO7ljQhkp3x1XKWmHEN-3Yn0XVpz/s1600/Big+Data+Talent.JPG" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" height="286" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgsKGODCbgPRmCskubXbwyMktzhnC9rPVQSxk6uuJ-bq4k-uwSk0GRiEwwzZHIlb9Rx0iwRVoH4rlJLviUlpzD9SB3X3am4Yfxl3mt7t87_JGgc5ayKcO7ljQhkp3x1XKWmHEN-3Yn0XVpz/s400/Big+Data+Talent.JPG" width="400" />
</a></div>
<span style="font-family: Calibri;">Here is a fantastic <a href="http://www.mckinsey.com/en/features/big_data.aspx">interactive visualization</a> of the existing big data talent and expected talent flows over the coming years.</span>
<br />
<div class="MsoNormal" style="margin: 0in 0in 10pt;"><br />
<span style="font-family: Calibri;">
<i style="mso-bidi-font-style: normal;">“A shortage of people with the skills necessary to take advantage of the insights that large datasets generate is one of the most important constraints on an organization’s ability to capture the potential from big data. <u>Leading companies are already reporting challenges in hiring this type of talent</u>.”</i> <span style="font-size: xx-small;">(103)</span> This shortage will be particularly acute for <i style="mso-bidi-font-style: normal;">“people with deep expertise in <a href="http://en.wikipedia.org/wiki/Machine_learning">statistics</a> and <a href="http://en.wikipedia.org/wiki/Machine_learning">machine learning</a>, and the managers and analysts who know how to operate companies by using insights from big data.”
<span style="font-size: xx-small;">(10)
</span></i> None of these talent gaps can be filled quickly, or addressed without a training pipeline. Even in industries which can attract talent, experiential industry-specific learning is needed to supplement the requisite big data skills… a process which takes a long investment of time.<o:p></o:p></span></div><br />
<b style="mso-bidi-font-weight: normal;"><span style="font-size: 14pt; line-height: 115%;">
<h1>
<span style="font-family: Calibri;">Quantification of Benefits</span>
</h1>
</span></b><br />
<span style="font-family: Symbol; mso-bidi-font-family: Symbol; mso-fareast-font-family: Symbol;">
<span style="mso-list: Ignore;">·<span style="font-size-adjust: none; font-stretch: normal; font: 7pt/normal "Times New Roman";"> </span></span></span><span style="font-family: Calibri;">The US healthcare system could save <em>“$300 billion in value every year”, including a reduction in “national health care expenditures by about 8 percent.”</em> <span style="font-size: xx-small;">(2)<o:p></o:p></span></span><br />
<span style="font-family: Symbol; mso-bidi-font-family: Symbol; mso-fareast-font-family: Symbol;"><span style="mso-list: Ignore;">·<span style="font-size-adjust: none; font-stretch: normal; font: 7pt/normal "Times New Roman";"> </span></span></span><span style="font-family: Calibri;">McKinsey estimates that a retailer fully leveraging big data can <em>“increase its operating margin by more than 60 percent.”</em> <span style="font-size: xx-small;">(2)<o:p></o:p></span></span>
<br />
<span style="font-family: Symbol; mso-bidi-font-family: Symbol; mso-fareast-font-family: Symbol;">
<span style="mso-list: Ignore;">·
<span style="font-size-adjust: none; font-stretch: normal; font: 7pt/normal "Times New Roman";">
</span>
</span>
</span>
<span style="font-family: Calibri;">McKinsey estimates that western European governments can save <em>“$150 billion in operational efficiency improvements alone by using big data.”</em> The value is much greater if the error elimination, anti-fraud efforts, and tax evasion applications of big data are considered. <span style="font-size: xx-small;">(2)
</span>
</span>
<br />
<span style="font-family: Symbol; mso-bidi-font-family: Symbol; mso-fareast-font-family: Symbol;"><span style="mso-list: Ignore;">·
<span style="font-size-adjust: none; font-stretch: normal; font: 7pt/normal "Times New Roman";"> </span></span>
</span>
<span style="font-family: Calibri;">Accumulating scarce talent in advance of demand positions a company well and gives a first mover advantage.</span><br />
<span style="font-family: Symbol; mso-bidi-font-family: Symbol; mso-fareast-font-family: Symbol;"><span style="mso-list: Ignore;">·<span style="font-size-adjust: none; font-stretch: normal; font: 7pt/normal "Times New Roman";"> </span></span></span><span style="font-family: Calibri;">To the extent that analytics are patentable, competitors are forced to use less attractive solutions for decades.</span><br />
<br />
<b style="mso-bidi-font-weight: normal;"><span style="font-size: 14pt; line-height: 115%;">
<h1>
<span style="font-family: Calibri;">Commentary
</h1>
</span></span></b><br />
<br />
<span style="font-family: Calibri;">DATA INSPIRED MERGERS:<strong><em> </em></strong>“Consolidation in an industry can also bring about beneficial scale effects in the aggregation and analysis of data.” <span style="font-size: xx-small;">(101)<o:p></o:p></span></span><br />
<br />
<span style="font-family: Calibri;">DATA IS THE FOUNDATION OF CAvQM:<strong> </strong> I stated “Data is the foundation of CAvQM.” In my Cemex blog, and McKinsey apparently agrees, “Access to big data is a key prerequisite to capturing value.” <span style="font-size: xx-small;">(102)</span></span><br />
<br />
<span style="font-family: Calibri;">FACTOR OF PRODUCTION: I agree that data is a factor of production, but I don’t agree that this is a recent development. Cemex was doing this in the 90’s, and the <a href="http://www.neilsonjournals.com/OMER/sOMTriangle.pdf">OM triangle</a> has always pointed to a trade- off between information and capacity… which is inextricably linked to labor and capital.<o:p></o:p></span><br />
<br />
GREATER INCOME DISPARITY: When productivity explodes it will make big data skills more valuable, and the shortage of talent will drive compensation up further.<br />
<br />
<br />
<br />
<span style="font-size: x-small;">"Big data: The next frontier for innovation, competition, and productivity." McKinsey Global Institute. May 2011. By James Manyika, Michael Chui, Brad Brown, Jacques Bughin, Richard Dobbs, Charles Roxburgh, Angela Hung Byers.</span><br />
<br />
<span style="font-size: x-small;">"Business Schools Plan Leap Into Data." Wall Street Journal. 04AUG2011. Melissa Korn and Shara Tibken.</span></div>CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.comtag:blogger.com,1999:blog-4323599490342956532.post-80526111193416482022011-09-03T19:13:00.000-04:002011-12-18T01:29:37.878-05:00Predicting Pitcher Injuries with Neural Networks<div class="MsoNormal">My <a href="http://cavqm.blogspot.com/2010/08/moneyball-art-of-winning-unfair-game-by.html">first blog entry</a> documented the Oakland A’s quantitative strategy as described in the book <a href="http://en.wikipedia.org/wiki/Moneyball"><u>Moneyball</u></a>. Although seeking more novel applications of QM, I recently stumbled on <a href="http://en.wikipedia.org/wiki/Moneyball"><u>The Extra 2%: How Wall Street Strategies Took a Major League Baseball Team from Worst to First</u></a> which includes a chapter on the quantitative strategy deployed by the Tampa Bay Rays, to be examined below.</div><div class="MsoNormal"><br />
</div>
<div class="MsoNormal">
<h1>
<b>Pitchers Are Expensive</b>
</h1>
</div>
<div class="MsoNormal">By far the most interesting application of QM by the Rays is predicting pitcher injuries. Because pitchers are expensive and significantly influence the team's defense, Tampa Bay sought a way to predict and thereby prevent pitcher injuries. This effort was targeted at pattern detection in the final ten throws preceding a disabling injury, which was fed into an algorithm along with countless other data points. The Rays ultimately tasked an employee with the absolute power to pull a pitcher if the algorithm predicted an injury in the immediate future. While this is premised largely on the idea that fatigue and over-exertion contribute to injuries, this has been confirmed by other pitching analysis such as Pitch F/X.</div><div class="MsoNormal"><br />
</div><div class="MsoNormal">
<h1>
<b>Pitch F/X</b>
</h1>
</div>
<div class="MsoNormal">The path to pitcher injury prediction was opened when MLB.com implemented the <a href="http://mlb.mlb.com/news/article.jsp?ymd=20071002&content_id=2245402&vkey=news_mlb&fext=.jsp&c_id=mlb">Pitch F/X</a> system in 2007 to precisely track and record data points on every pitch. </div>
<div class="MsoNormal"><br />
</div>
<div class="MsoNormal">
</div>
<div class="MsoNormal">
<span class="apple-style-span">
<i>
<span style="background: none repeat scroll 0% 0% white; color: black; font-family: "Verdana","sans-serif"; font-size: 9pt; line-height: 115%;">“
<a href="http://www.slate.com/id/2172223/">Now the computer generates all of this [data] automatically</a>—how high the pitcher's throwing hand was off the ground when he released the ball, how fast the ball was traveling both when it left his hand</span>
</i></span><span class="apple-converted-space"><i><span style="background: none repeat scroll 0% 0% white; color: black; font-family: "Verdana","sans-serif"; font-size: 9pt; line-height: 115%;"> </span>
</i></span>
<em>
<span style="background: none repeat scroll 0% 0% white; color: black; font-family: "Verdana","sans-serif"; font-size: 9pt; font-style: normal; line-height: 115%;">and</span>
</em><span class="apple-converted-space"><i><span style="background: none repeat scroll 0% 0% white; color: black; font-family: "Verdana","sans-serif"; font-size: 9pt; line-height: 115%;"> </span>
</i></span>
<span class="apple-style-span">
<i>
<span style="background: none repeat scroll 0% 0% white; color: black; font-family: "Verdana","sans-serif"; font-size: 9pt; line-height: 115%;">when it crossed the plate, to what degree and in what direction the ball diverted from a straight path on its way to the plate, and finally, if the pitch really was four inches inside and a couple of inches above the knees.”</span>
</i>
</span>
</div><div class="MsoNormal">
<a href="http://www.slate.com/id/2172223/"></a></div>
<div class="MsoNormal"><br />
</div>
<div class="MsoNormal">This wealth of data points as well as 'arm slot', 'velocity' and injury data feeds a neural network <a href="http://en.wikipedia.org/wiki/Artificial_neural_network">(a non-linear algorithm that loosely imitates brain function)</a> trained to identify tell-tale signs of pitcher injury.While injury prevention is the most interesting application of QM (and a very profitable one), there were other analyses with profound implications for the team.</div><div class="MsoNormal"><br />
</div>
<div class="MsoNormal">
<h1>
<b>Extra Bases Run</b>
</h1>
</div>
<div class="MsoNormal">Tampa Bay also applied QM to their base running strategy. <span> </span>This involved training players to keep running if fielders didn’t have the ball. A simple heuristic to be sure, but a non-standard one in Major League Baseball and one which propelled them from 25<sup>th</sup> to 1<sup>st</sup> in extra bases run. Given the potency of this rule, I’m confident that a statistician tested it using league data before Tampa Bay implemented it.</div><div class="MsoNormal"><br />
</div>
<div class="MsoNormal">
<h1>
<b>Statistical Justification of Onfield Strategy</b>
</h1>
</div>
<div class="MsoNormal"><u>Moneyball</u> detailed the first Major League efforts to debunk baseball tactics such as sacrifice bunts and stealing bases although <u>The Extra 2%</u> offers several of its own. These include using extra in-fielders for batters who frequently hit ground balls, extra out-fielders for batters with a tendency for fly outs, shifting the infield left or right depending on a batters ground out tendencies, and an extra fielder in center right against Derek Jeter.</div>
<div class="MsoNormal"><i>“Maddon, whose more adventurous against-the-Book decisions included issuing the infamous bases-loaded intentional walk to Josh Hamilton, sending a runner while down 9-0, using unusually aggressive defensive shifts, and starting same-hand hitters against a handful of quirky pitchers.”</i><span> </span>(Page 172)</div>
<div class="MsoNormal"><br />
</div>
<div class="MsoNormal">
<h1>
<b>Quantification of Benefits</b>
</h1>
</div>
<ul style="text-align: left;"><li><span style="font-family: Symbol;"><span><span style="font: 7pt "Times New Roman";"></span></span></span>Won the American League pennant</li>
<li><span style="font-family: Symbol;"><span><span style="font: 7pt "Times New Roman";"></span></span></span>Went from 25<sup>th</sup> to 1<sup>st</sup> in the league for extra bases run</li>
<li><span style="font-family: Symbol;"><span><span style="font: 7pt "Times New Roman";"></span></span></span>Greater pitcher ROI and utilization which reduces rotation of non-injured pitchers</li>
<li><span style="font-family: Symbol;"><span><span style="font: 7pt "Times New Roman";"></span></span></span>Pitcher Arbitrage</li>
<li>Similar to the Oakland A's, they can compete against the Yankees with a third the player salaries</li>
</ul><div class="MsoNormal"><br />
</div>
<div class="MsoNormal"> <b>Personal Commentary</b></div>
<div class="MsoNormal"><br />
</div>
<div class="MsoNormal"><i>Confidentiality of Methodology:</i> The management of this team knows what they’re doing. For example, when they hired the Pitcher injury prediction specialist, they hid it from the world by not adding his name to the company directory, not letting him announce why he was closing his blog, etc.</div><div class="MsoNormal"><br />
</div>
<div class="MsoNormal"><i>Injured Pitcher Arbitrage:</i> I’m especially interested though in the implications of the pitcher injury prediction. More specifically, a team with this capability can profitably arbitrage injury prone pitchers. Familiarity with league-wide injuries as well as the additional information known from the injury-prone pitcher’s history would produce a robust injury prevention capability. After all, to what extent is an injured pitcher the result of poor management vs. poor self-knowledge?</div><div class="MsoNormal"><br />
</div>
<div class="MsoNormal"><i>Data Cleansing as Competitive Advantage: </i>Several <a href="http://www.beyondtheboxscore.com/2011/2/10/1982529/being-cautious-with-using-pitchf-x-data-to-evaluate-stuff-the-case-of">articles</a> mention calibration problems with the Pitch F/X cameras, leading to different baselines in different stadiums. Could an analytically oriented team gain an advantage over other analytical teams by deliberately skewing the cameras in their own stadium, and by excelling in their normalization or data cleansing methods?</div><div class="MsoNormal"><br />
</div><span style="font-size: large;"><i><b>The automated collection and public distribution of data creates an inflection point.</b></i></span></div>CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.comtag:blogger.com,1999:blog-4323599490342956532.post-34822708718559684052011-07-10T01:22:00.002-04:002012-03-28T22:09:04.751-04:00Reverse Engineering The IRS DIF-Score<div dir="ltr" style="text-align: left;" trbidi="on">
<a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiVzJG10NR23_OM5TQi0ozXB0Rjm_O4Fqqg08rTNtDgUDxJOwoAd9vKhnjEuimpSQ2iZqXMzn7NvC2vaWHaPRLnQGAnTmzx9Ah7OlWyv70G_EJbgJ9MI11J6QVXfe8DzSXagdc_D8vHOBZ3/s1600/IRS.jpg" imageanchor="1" style="clear: right; float: right; margin-bottom: 1em; margin-left: 1em;">
<img border="0" height="260" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiVzJG10NR23_OM5TQi0ozXB0Rjm_O4Fqqg08rTNtDgUDxJOwoAd9vKhnjEuimpSQ2iZqXMzn7NvC2vaWHaPRLnQGAnTmzx9Ah7OlWyv70G_EJbgJ9MI11J6QVXfe8DzSXagdc_D8vHOBZ3/s320/IRS.jpg" width="320" />
</a>
<br />
<div>
What would you do if you were a statistician audited by the IRS?
<a href="http://www.nytimes.com/1996/02/25/business/your-taxes-some-new-tricks-to-help-filers-avoid-an-old-audit-trap.html?pagewanted=all&src=pm">Dr. Amir Aczel</a>, Professor of Statistics at Bentley college and author of the 1995 book <a href="http://www.loompanics.com/TaxAvoision/tax_book1.htm">How To Beat The IRS At Its Own Game</a> chose to reverse engineer the IRS' <a href="http://www.irs.gov/pub/irs-soi/puidif2.pdf">secret audit selection formula</a> (the <a href="http://www.fool.com/personal-finance/taxes/2006/03/17/top-5-audit-myths.aspx"><span>DIF score</span></a>) using quantitative methods and publish it to the world.<br />
<br />
<h1>
The Secret Formula: How The IRS Picks Audits</h1>
<br />
The IRS uses a quantitative method called <a href="http://en.wikipedia.org/wiki/Linear_discriminant_analysis">Discriminant Analysis</a> to identify the 'underreporters' from the normal returns, driven largely by the following details.<br />
<ul style="text-align: left;">
<li><i>Schedule A Ratio:</i> They'll audit you if your schedule A (Itemized) deductions are more than 44% of your income.</li>
<li><i>Schedule C Ratio:</i> They'll audit you if your ratio of schedule C (Business) deductions is more than 63% of income.</li>
<li><i>Schedule F Ratio:</i> They'll audit you if your ratio of schedule F (Farm) deductions is more than 67% of income.</li>
<li>Audits are 4x more likely if you own a business and 2x if you own a farm.</li>
<li>The Obama administration has focused on high earners: </li>
<ul>
<li>Audits <a href="http://money.cnn.com/2005/01/31/pf/taxes/avoid_audit/index.htm">5x more likely</a> if your <i>income is above $100,000.</i></li>
<li><a href="http://blogs.wsj.com/wealth/2011/03/22/irs-targets-rich-taxpayers-audit-rates-up-80/">20% chance of audit</a> if you make $10+ million. (20x the average audit rate)</li>
<li><a href="http://blogs.wsj.com/wealth/2011/03/22/irs-targets-rich-taxpayers-audit-rates-up-80/">12% chance of audit</a> if you make $5-10 million. (12x the average audit rate)</li>
</ul>
<li><i>Occupation affects your audit likelihood</i> </li>
<ul>
<li>22% of business and personal services companies are audited every year.</li>
<li>16% of building contractors are audited every year.</li>
</ul>
<li>Returns filed <i>after</i> April 15 aren't audited. </li>
</ul>
<b>
<br />
</b>
<br />
<h1>
Debunking Myths</h1>
<br />
Dr. Aczel refuted a number of IRS myths in addition to reverse engineering the IRS audit formula.<br />
<ul style="text-align: left;">
<li><i>Debunked IRS Omniscience:</i> The IRS previously held that the formula was complex and included countless variables known about your life.</li>
<li><i>Debunked Annual <span>DIF Score</span> Updates: </i>The IRS formerly stated that they modify the selection algorithm yearly but Aczel could find no detectable changes. </li>
<li><i>Highlighted Politically Motivations Rather Than Economic: </i>At the time of writing, Congress was considering budget cuts to the IRS. The Commissioner tried to prove the agency's value by tripling the number of audits, albeit by focusing on lower payout cases.</li>
</ul>
<h1>
Game Theoretic Interpretation of the Audit Process</h1>
<br />
<div>
<ul style="text-align: left;">
<li><b>Entrapment:</b> Before reviewing the case, the Auditor may offer to settle for half of your claimed deductions. They are trained to make this offer for two reasons:</li>
<ul>
<li>You may settle solely to avoid the hassle, but the IRS will interpret this as an implicit admission of guilt. Accordingly, they will audit you repeatedly throughout your life.</li>
<li>Following the case to completion will halve your tax obligation (at <i>least</i>). So the auditor doesn't sacrifice anything by making this offer.</li>
</ul>
<li><b>Auditors Don't Want Cases With Taxpayers That Appeal:</b> IRS agents select cases from a pool chosen by the <span>DIF Score</span>.<i> "The auditor's performance is judged by the IRS based on the number of cases closed with the taxpayer's agreement."</i> If you settle quickly and pay a fine, other agents will gladly audit you in the future. Your best deterrent is to fight for every deduction, let the process fully run its course, escalate it to a supervisor and use the appeals process because it makes you an unattractive target for future years.</li>
<li><b>Always Appeal. (85% Settled With An Average 40% Reduction In Tax)</b> Always appeal the Senior Auditor's decision. The appeals officer is <i>'judged by how many cases he or she settles with the taxper, rather than on backing up the original auditor's report.'</i></li>
<li><b>Post-Appeal. (<i>Another</i> 85% Are Settled Pre-Trial)</b> Even <i>after</i> the appeals process, 85% of the <i>post</i>-appeals cases are closed before they get to trial. Why? The IRS' 'legal resources are scarce'.</li>
</ul>
</div>
<h1>
Commentary</h1>
<br />
<div>
I love how the learnings completely reverse the negotiation. Firstly, taxpayers are empowered to confront the auditor with justifications for the item which flagged them. Secondly, this information empowers taxpayers to avoid audits by delaying their tax filing until August, rebalancing deductions to different Schedules or by providing thorough explanations for items that will be flagged by the IRS. Lastly, QM removes the myth of omniscience and proves that the audit payoff is counter-intuitive, because you save more money the longer you continue the process (despite any comments from the auditor). It is probably the <i>only</i> example in government relations where you should ignore their recommendations (they're not mandates) and just consistently escalate the issue.<br />
<br />
These findings are also disconcerting because they point to an ineffective detection mechanism for tax evasion. It means that someone with no legitimate deductions but who claims 20% <i>won't</i> be audited, while someone with a legitimately high number of claims <i>will</i> be audited. Looking at this from the systems perspective, the IRS is probably chasing the same group of legitimate high-deducters year after year while completely missing the filers with smaller illegitimate claims. One such example was mentioned:<br />
<i><span class="Apple-style-span" style="font-size: x-small;">"During one of my radio interviews, on WJJD in Chicago, a lady called in and described on the air how every single year she is audited by the IRS. She and her husband, apparently, make a modest living, and don't file any 'complicated' tax forms. They don't own a business or a farm; their charitable contributions are moderate; they don't have a home office. But they have 21 children. She described how every year her husband has to go downtown to meet with the IRS after receiving the letter asking them to come for an examination."</span></i><br />
Do they really need to validate this <i>every</i> year? Wouldn't they eventually realize that the dependents are all verified?<br />
<br />
Lastly, am I the only one shocked at the level of subjectivity in these proceedings? <i><span class="Apple-style-span" style="font-size: x-small;">"Every year, Money magazine asks a group of 50 top tax experts to prepare the return of a hypothetical individual, and they then publish the comparison of the experts' results. In all the years this has been don, not a single time have even two experts agreed on the exact amount of tax due!"</span></i><br />
<br />
<br /></div>
<u>How to Beat The I.R.S. At Its Own Game.</u> By Amir Aczel, Ph.D. Four Walls Eight Windows Publishing, 1995. New York, NY.<br />
<br />
<a href="http://www.nytimes.com/2010/03/13/business/13tax.html">http://www.nytimes.com/2010/03/13/business/13tax.html</a><br />
<br />
<a href="http://www.nytimes.com/1996/02/25/business/your-taxes-some-new-tricks-to-help-filers-avoid-an-old-audit-trap.html?pagewanted=all&src=pm">http://www.nytimes.com/1996/02/25/business/your-taxes-some-new-tricks-to-help-filers-avoid-an-old-audit-trap.html?pagewanted=all&src=pm</a><br />
<br />
http://www.fool.com/personal-finance/taxes/2006/03/17/top-5-audit-myths.aspx<br />
<br />
<a href="http://money.cnn.com/2005/01/31/pf/taxes/avoid_audit/index.htm">http://money.cnn.com/2005/01/31/pf/taxes/avoid_audit/index.htm</a><br />
<br />
<a href="http://blogs.wsj.com/wealth/2011/03/22/irs-targets-rich-taxpayers-audit-rates-up-80/">http://blogs.wsj.com/wealth/2011/03/22/irs-targets-rich-taxpayers-audit-rates-up-80/</a><br />
<br />
<a href="http://www.bankrate.com/brm/itax/news/20030221a2.asp">http://www.bankrate.com/brm/itax/news/20030221a2.asp</a><br />
<br />
<a href="http://www.irs.gov/pub/irs-soi/puidif2.pdf">"Predictors Of Unreported Income: Tests of Unreported Income (UI) <span>DIF Scores.</span>"</a> by Dennis Cyr, Thomas Eckhardt, Lou Ann Sandoval, and Marvin Halldorson. Internal Revenue Service Research Conference.<br />
June 11, 2002.</div>
</div>CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.comtag:blogger.com,1999:blog-4323599490342956532.post-28420283958558740772011-05-30T22:53:00.000-04:002011-12-17T12:46:33.155-05:00Evolutionary Police Sketch Software and Genetic Algorithms<a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiJDaD6Cmdph3rS58nwXqnsVMssxPmkM-eWhJohkpH86nvQIWS1X6piOzZCA_AzuPFRK4d2sGMJb7umnv4SWENVpw2-yZmBRMt60mASujQuSMnEhlYGzxlQvAigmpMiQJK3c6pDkh77ZgHh/s1600/IQ+Biometrix+Software.png" imageanchor="1" style="clear: right; cssfloat: right; float: right; margin-bottom: 1em; margin-left: 1em;"><img border="0" height="249" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiJDaD6Cmdph3rS58nwXqnsVMssxPmkM-eWhJohkpH86nvQIWS1X6piOzZCA_AzuPFRK4d2sGMJb7umnv4SWENVpw2-yZmBRMt60mASujQuSMnEhlYGzxlQvAigmpMiQJK3c6pDkh77ZgHh/s320/IQ+Biometrix+Software.png" t8="true" width="320" /></a>The <a href="http://en.wikipedia.org/wiki/Facial_composite">FBI prefers</a> artists for sketching suspects, but software is rapidly making inroads. Demand for a software solution was created by the many police departments that couldn't afford such an artist.<br />
<br />
The first generation of sketch software, such as <a href="http://www.facesid.com/products_faces_demo.html">FACES</a>, asked victims to select individual features from a catalogue, as shown at the right.
<br />
Recently developed software avoids the catalogue by using 'evolutionary' techniques. <a href="http://www.evofit.co.uk/">EvoFIT</a> asks users to click on the 6 photos out of 70 that look the most like the assailant. The six chosen faces are then combined using genetic algorithms to create 70 offspring images. The user continues this process until satisfied with the criminal's likeness.<br />
<br />
<h1>Evolutionary Algorithms</h1>
<br />
The novelty in the EvoFIT approach is the combination of multiple pictures to produce various offspring, which is done with genetic algorithms. <a href="http://en.wikipedia.org/wiki/Genetic_algorithm">Genetic algorithms</a> imitate the reproductive process by blending characteristics from two patterns or parents ('parents' could be images, a job rotation schedule, chess simulation, or investment strategies) into multiple 'offspring'. This cycle is repeated until a satisfactory answer is found, but the interesting point is that the offspring are usually superior to the parents (this is why evolution works). In this particular instance, 'superior' means they're more like the actual criminal than the parent images.<br />
<br />
<h1>A Picture is Worth A Thousand Words</h1>
<br />
In this example, users were asked to<a href="http://books.google.com/books?id=fI_LI6a-e3IC&pg=PA32&dq=the+darwinian+police+sketch&hl=en&ei=8P_jTbHCF-rh0QG-sY2xBw&sa=X&oi=book_result&ct=result&resnum=1&ved=0CC4Q6AEwAA#v=onepage&q=the%20darwinian%20police%20sketch&f=false"> 'sketch' Ben Affleck</a> with EvoFIT. The first picture is one of the initial 6 chosen by the user. The second picture is one of the 70 offspring from the first generation, and the third picture is one of the offspring from the second generation of likenesses. <a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhQJePhr1pzgc3YanI-alMk4iy9IHq4e_0I0IxyDc5Ir5oG4I1jgntIBKw9wQx2GeXqv-IEJWmUXBfzdLv7rxHGXZsB0RcC1N5X7CeyJZevvEhsIml0LTJ-Zri1Kescx5v348FqzVWWqsLy/s1600/Affleck.jpg" imageanchor="1" style="clear: right; cssfloat: right; float: right; margin-bottom: 1em; margin-left: 1em;"><img border="0" height="200" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhQJePhr1pzgc3YanI-alMk4iy9IHq4e_0I0IxyDc5Ir5oG4I1jgntIBKw9wQx2GeXqv-IEJWmUXBfzdLv7rxHGXZsB0RcC1N5X7CeyJZevvEhsIml0LTJ-Zri1Kescx5v348FqzVWWqsLy/s200/Affleck.jpg" t8="true" width="161" /></a><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjjgb38s_CRkyJ2wPVP2Gj2p7eifPKAvb19DrSavwAbITO-zvXTaNY515xO_r2HPXgTOmoAyxXfOwhmAAYrTwgEBVVEBZgLpZts48IT9kjHZ8nglsX1-WpYk98rwtza2Wf76prnk6e0E1lL/s1600/darwinian+Affleck+Comparison.jpg" imageanchor="1" style="clear: left; cssfloat: left; float: left; margin-bottom: 1em; margin-right: 1em;"><img border="0" height="166" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjjgb38s_CRkyJ2wPVP2Gj2p7eifPKAvb19DrSavwAbITO-zvXTaNY515xO_r2HPXgTOmoAyxXfOwhmAAYrTwgEBVVEBZgLpZts48IT9kjHZ8nglsX1-WpYk98rwtza2Wf76prnk6e0E1lL/s320/darwinian+Affleck+Comparison.jpg" t8="true" width="320" /></a><br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
Note how the nose and brow are consistent over the generations, while the eyes and chin grow more similar to Ben Affleck and the shape becomes more oblong. I also think that the middle picture looks like Ben Affleck's brother <a href="http://en.wikipedia.org/wiki/Casey_Affleck">Casey Affleck</a>, which further demonstrates the parrallels between the algorithm and real world genetic similarity.<br />
<br />
<h1>Automated Picture Matching </h1>
<br />
Witnesses often look through hundreds of criminals' pictures in hopes of catching them. Witnesses grow tired after looking at this number of photos though, and it impedes their ability to recognize the assailant. Furthermore, our perceptions over-emphasize weight gain, hair color, or aging of the assailant compared to when the picture was taken. Accordingly, <a href="http://www.gizmag.com/software-developed-to-match-police-sketches-to-mug-shots/18060/">computers are now used to match sketches</a> to a database of mug shots, because it works instantaneously and it relies on unalterable relationships in facial structure. More specifically, this software examines the location of the nose relative to the mouth and eyes, length of mouth, width of nose, length of nose, location of eyebrow ridges, etc.<br />
<br />
<h1>Quantification and Benefits: Sketching</h1>
<ul>
<li style="border-bottom: medium none; border-left: medium none; border-right: medium none; border-top: medium none;">Greatly improved the accuracy of police sketches. </li>
<li style="border-bottom: medium none; border-left: medium none; border-right: medium none; border-top: medium none;">Yields a 45% arrest rate vs. single digits before EvoFIT was used.</li>
<li style="border-bottom: medium none; border-left: medium none; border-right: medium none; border-top: medium none;"><em>"</em><a href="http://www.popsci.com/gear-gadgets/article/2005-06/darwinian-police-sketch"><em>victims no longer</em></a><em> have to worry about describing their assailant."</em></li>
<li style="border-bottom: medium none; border-left: medium none; border-right: medium none; border-top: medium none;"><em>"The software exploits the fact that the human brain is much better at recognizing a face than describing it."</em></li>
<li><em>"we can </em><a href="sketching: "we can schedule (the interview) around the victim's time rather than the artist's time""><em>schedule</em></a><em> (the interview) around the victim's time rather than the artist's time"</em></li>
<li>Rapid, iterative creation of sketches. Approximately half the time taken with artist sketches.</li>
</ul>
<h1>Quantification and Benefits: Matching</h1>
<ul>
<li><a href="http://www.gizmag.com/software-developed-to-match-police-sketches-to-mug-shots/18060/">45% match</a> from the sketch to the arrestant. </li>
<li>Instantaneous, so the suspect has less time to flee.</li>
<li>Overcomes witness fatigue, because they don't need to look at hundreds of mug shots.</li>
<li>Reduces the time required by the witness.</li>
<li>Protects the innocent. Would you rather be one of 6 in a lineup or one person in a database of a million?</li>
</ul>
<br />
<br />
<h1>Commentary</h1>
<u>If you can't identify a suspect, then you can't arrest them.</u> The drastic improvement in victim's facial recall and sketch matching makes the entire criminal justice system work better. While I am sure that some criminals are convicted based on CSI results, I think the ability to identify a suspect is <u>critical</u> to the entire system. For example, if you were a fisherman, would you rather have a fish on the hook 3% or 90% of the time?<br />
<br />
<a href="http://en.wikipedia.org/wiki/Facial_composite">
</a>
<br />
<br />
<span style="font-size: x-small;">"Sketch Software Still Suspect." June 2000.
</span>
<a href="http://www.wired.com/science/discoveries/news/2000/06/37122">
<span style="font-size: x-small;">http://www.wired.com/science/discoveries/news/2000/06/37122
</span>
</a>
<br />
<span style="font-size: x-small;">Interquest Faces Software
</span>
<a href="http://www.iqbiometrix.com/tech_support_overview.html">
<span style="font-size: x-small;">http://www.iqbiometrix.com/tech_support_overview.html
</span>
</a>
<br />
<span style="font-size: x-small;">"The Darwinian Police Sketch." Popular Science, July 2005.
</span>
<br />
<a href="http://www.gizmag.com/software-developed-to-match-police-sketches-to-mug-shots/18060/">
<span style="font-size: x-small;">http://www.gizmag.com/software-developed-to-match-police-sketches-to-mug-shots/18060/
</span>
</a>
<a href="http://www.facesid.com/products_faces_demo.html">
<span style="font-size: x-small;">http://www.facesid.com/products_faces_demo.html
</span>
</a>CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.comtag:blogger.com,1999:blog-4323599490342956532.post-57449975552565637642011-05-09T23:34:00.000-04:002011-12-17T12:26:45.167-05:00The $1,000,000 Netflix Challenge<div style="border-bottom: medium none; border-left: medium none; border-right: medium none; border-top: medium none;"><a href="http://www.netflix.com/" imageanchor="1" style="clear: right; cssfloat: right; float: right; margin-bottom: 1em; margin-left: 1em;"><img border="0" height="92" j8="true" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEh8vZjoXhn2acLFdd7G4f4uZD54-zRIsTBSKHjmlKBufRIsxoU26uO77-1pFf83M4fDqLgYGiD_kD0uZDJf7fDtuxqQqkUSjpPkT9GpOJoeO7QK2HKd3POYaTNgheD8m_N2k-UwHsUuZGg7/s200/untitled.bmp" width="200" /></a>Netflix <em>" has more than 7 million subscribers, who have the option to rate movies on a scale of 1 to 5. To encourage users to keep their subscriptions active, Netflix rolled out Cinematch, which used those ratings to help customers find new movies they'd like. When a user logs in, the service suggests "Movies You'll Love" — a list of films that the algorithm guesses will get a high rating from that particular user." </em>To improve customer recommendations they offered a $1 million prize in 2006 to the first team to create an algorithm 10% more accurate than Cinematch. After thousands of hours of
<a href="http://en.wikipedia.org/wiki/Crowdsourcing">crowdsourced</a> labor were invested in this challenge, Netflix paid 7 statisticians, machine-learning experts and engineers
<a href="http://www.netflixprize.com/">$1 million dollars for an algorithm</a> on September 21, 2009. This may not surprise you after learning that they were led by AT+T Research engineers and worked on the algorithm for three years, but this does beg the question, <u>"What did they create and why was it so valuable?"</u>
</div><div style="border-bottom: medium none; border-left: medium none; border-right: medium none; border-top: medium none;"></div><br />
<h1>The Netflix Challenge</h1><br />
Netflix views their business as more than renting videos to customers, <em>" </em><a href="http://books.google.com/books?id=O2k0K1w_bJIC&printsec=frontcover&dq=the+long+tail&hl=en&ei=Lz-_TdjnNaXb0QHjuJG5BQ&sa=X&oi=book_result&ct=result&resnum=1&ved=0CCoQ6AEwAA#v=snippet&q=netflix%20blockbuster&f=false"><em>we create demand for content</em></a><em> and we help you find great movies that you'll really like." </em>Given this business model they created the <a href="http://en.wikipedia.org/wiki/Netflix_Prize">Netflix Challenge</a>. The competition began with hundreds of competitors trying their hand at recommendation system algorithms. After a year, progress slowed to a crawl with entries only halfway to their goal, when an unemployed psychologist/engineer quickly rose to the top 5 by <a href="http://www.wired.com/techbiz/media/magazine/16-03/mf_netflix">using psychological principles to improve his algorithm</a>. After little improvement for months, the top two teams combined and cleared the winning threshold. This triggered a 30-day scramble by other teams to beat the winning algorithm prior to contest closure. Thirty lower ranked teams then combined to become 'The Ensemble' and turn in <em>better</em> results than the top team. <em>"By pooling a larger number of competitors — all of whose algorithms performed more poorly than those of the top two teams — The Ensemble produced the best results. It’s a profound lesson in the power of the crowd." </em>The Ensemble ultimately lost due to the award criteria, but the competition is rife with strategic implications that will reverberate for years.<br />
<br />
<h1>Data Mining Insights (How The Teams Created Better Recommendations)</h1><br />
<ul>
<li><strong>Temporal Effects:</strong> <em>"viewers in general tend to <a href="http://www.wired.com/epicenter/2009/09/how-the-netflix-prize-was-won/">rate movies differently on Fridays versus Mondays</a>, and certain users are in good moods on Sundays, and so on."</em></li>
<li><strong>Rating Immediately vs. By Memory:</strong> <em>"As it turns out, people who rate a whole slew of movies at one time tend to be rating movies they saw a long time ago. The data showed that <a href="http://www.wired.com/epicenter/2009/09/how-the-netflix-prize-was-won/">people employ different criteria to rate movies they saw a long time ago</a>, as opposed to ones they saw recently — and that in addition, some movies age better than others, skewing either up or down over time."</em></li>
<li><strong>Anchoring:</strong> <em>"If a customer watches three movies in a row that merit four stars — say, the <cite>Star Wars</cite> trilogy — and then sees one that's a bit better — say, <cite>Blade Runner</cite> — they'll likely give the last movie five stars. But if they started the week with one-star stinkers like the <cite>Star Wars</cite> prequels, <cite>Blade Runner</cite> might get only a 4 or even a 3. Anchoring suggests that rating systems need to take account of inertia — a user who has recently given a lot of above-average ratings is likely to continue to do so."</em></li>
<li><strong>Customer Profiling:</strong> Classifying viewer preferences into "low brow" and "high brow", or by languages spoken, preference for modern movies or classics, etc.</li>
<div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgxm0RdYGcYKKbFFqRGfWXgA5vJ5IJRSk_M2cvPlM5hOVDmt5as6dIoENk5fiPSpOWoW08hkhvOAh4N1C2c5LceK-xAL_zrhFXHqvdz8j9tt4sK6muiVrDnFbksSYMAunDK5qqko_iado4V/s1600/Netflix+Similarity.jpg" imageanchor="1" style="clear: right; cssfloat: right; float: right; margin-bottom: 1em; margin-left: 1em;"><img border="0" height="320" j8="true" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgxm0RdYGcYKKbFFqRGfWXgA5vJ5IJRSk_M2cvPlM5hOVDmt5as6dIoENk5fiPSpOWoW08hkhvOAh4N1C2c5LceK-xAL_zrhFXHqvdz8j9tt4sK6muiVrDnFbksSYMAunDK5qqko_iado4V/s320/Netflix+Similarity.jpg" width="303" /></a></div>
<li><strong>Nearest Neighbor Classification: </strong>People like movies that are similar to other movies they like. (<a href="http://www.the-ensemble.com/content/netflix-prize-movie-similarity-visualization">Visualization of 5000 movies provided</a>)</li>
<li><strong>Weight Recent Rankings and Movie Selections More Heavily:</strong> Recent ratings are better indications of tastes than old ratings. For example, movies watched during college may not reflect your tastes as a young professional.</li>
</ul><strong>Benefits and Quantification</strong><br />
<span style="font-family: "Times New Roman", "serif"; font-size: 12pt; line-height: 115%; mso-ansi-language: EN-US; mso-bidi-language: AR-SA; mso-fareast-font-family: "Times New Roman"; mso-fareast-language: EN-US;"><em>"Netflix has said that winning <a href="http://www.wired.com/epicenter/2009/06/winning-teams-join-to-qualify-for-1-million-netflix-prize/">a 10 percent improvement on its recommendation algorithm for $1 million would be a tremendous bargain.</a>"</em></span><br />
<ul>
<li><strong>Lock-in Effect: </strong>After a customer has rated a thousand movies, Netflix is generating finely tuned recommendations and doesn't recommend previously watched movies. Will the user reenter those ratings on a different website that can't generate better results?</li>
<li><strong>Generates Better Return on DVD Inventory:</strong> From <u><a href="http://www.amazon.com/Long-Tail-Future-Business-Selling/dp/1401302378">The Long Tail</a></u> <em>"Historically, Blockbuster has reported that about 90% of the movies they rent are new theatrical releases. Online they're more niche: about 70% of what they rent from their website is new releases and about 30% is back catalog. That's not true for Netflix. About 30% of what [Netflix] rents is new releases and about 70% is back catalog and its not because we have a different subscriber." </em>This allows Netflix to advertise unused inventory to generate returns on existing investments. </li>
<li><strong>Reduces Stock Outs and Long Wait Times: </strong>By spreading customer demand over a broader variety and number of titles, they eliminate spikes in customer demand for a single title and thereby avoid long times.</li>
<li><strong>Increases Revenues:</strong> Encourages household accounts to change to individual accounts so that their recommendations are personalized. Also encourages migration from the streaming subscription to the more expensive "streaming and snail mail" package. After all, what good are your insightful recommendations if you can't access the non-streaming ones?</li>
<li><strong>Customer Experience Enhancement:</strong> Never watch a bad movie again. Titles delivered to you.</li>
<li><strong>Customer Loyalty: </strong>Better recommendations means fewer customer defections to competitors (Blockbuster or Redbox).</li>
</ul><br />
<div class="MsoNormal" style="line-height: normal; margin: 0in 0in 10pt; mso-margin-bottom-alt: auto; mso-margin-top-alt: auto;"><strong>CrowdSourcing Commentary</strong><br />
This challenge may be<em><strong> the</strong></em> best proof ever of the <a href="http://en.wikipedia.org/wiki/The_Wisdom_of_Crowds">wisdom of crowds</a>. </div><div class="MsoNormal" style="line-height: normal; margin: 0in 0in 10pt; mso-margin-bottom-alt: auto; mso-margin-top-alt: auto;"><span style="font-family: "Times New Roman", "serif"; mso-fareast-font-family: "Times New Roman";"><em>"the Netflix Prize competition has proffered hard proof of a basic crowdsourcing concept: Better solutions come from </em>
<a href="http://www.wired.com/epicenter/2009/09/how-the-netflix-prize-was-won/">
<em>unorganized people who are allowed to organize organically</em>
</a><em>. But something else happened that wasn’t entirely expected: Teams that had it basically wrong — but for a few good ideas — made the difference when combined with teams which had it basically right, but couldn’t close the deal on their own." </em></span><span style="font-family: "Times New Roman", "serif"; mso-fareast-font-family: "Times New Roman";"><em>"Ir</em></span><span style="font-family: "Times New Roman", "serif"; mso-fareast-font-family: "Times New Roman";">
<em>onically, <a href="http://www.wired.com/epicenter/2009/09/how-the-netflix-prize-was-won/">the most outlying approaches</a> — the ones farthest away from the mainstream way to solve a given problem — proved most helpful towards the end of the contest, as the teams neared the summit."</em>
</span>
<br />
<span style="font-family: "Times New Roman", "serif"; font-size: 12pt; mso-fareast-font-family: "Times New Roman";"><br />
<h1>Team Calibre Commentary</h1><br />
One article pointed out that crowdsourcing attracts talent that is otherwise unavailable because they always have jobs. The rather formidable <a href="http://bits.blogs.nytimes.com/2009/06/26/and-the-winner-of-the-1-million-netflix-prize-probably-is/">winning team included</a> 3 Research Engineers from AT+T Labs, a Senior Scientist from Yahoo Research in Israel, two machine learning consultants from Austria, and two Canadian engineers from the University of Toronto. Ignoring this team's credentials though, remember that there were <em>thousands</em> of such teams competing from around the world. <br />
<br />
<h1>Intellectual Property Commentary</h1><br />
At first I was dumbfounded that Netflix <a href="http://www.wired.com/epicenter/2009/06/1-million-netflix-prize-so-close-they-can-taste-it/">only required a license to the algorithm</a> at the end of the competition. They didn't require all intellectual property relating to it! I later realized that if delivery was required of the winning team though, then winning second place (and avoiding the hand-it-over requirement) becomes the optimal strategy for participants. Accordingly, Netflix let participants retain the intellectual property, which then interested AT+T enough to dedicate three scientists for years.<br />
<br />
<br />
<br />
<span style="font-size: x-small;">"This Psychologist Might Outsmart the Math Brains Competing for the Netflix Prize." By Jordan Ellenberg. Wired Magazine Issue 16-03. 25FEB2008.</span><br />
<span style="font-size: x-small;"></span><br />
<span style="font-size: x-small;"> </span><span style="font-size: x-small;"><u>The Long Tail.</u> by Chris Anderson. Page 109. Copyright 2006. Hyperion Books, New York, NY.</span><span style="font-size: x-small;"><br />
</span><br />
<a href="http://en.wikipedia.org/wiki/Netflix_Prize"><span style="font-size: x-small;">http://en.wikipedia.org/wiki/Netflix_Prize</span></a><br />
<br />
<a href="http://www.netflixprize.com/"><span style="font-size: x-small;">www.NetflixPrize.com</span></a><br />
<br />
<a href="http://www.the-ensemble.com/"><span style="font-size: x-small;">http://www.the-ensemble.com/</span></a><br />
<br />
<span style="font-size: x-small;">"$1 Million Netflix Prize So Close, They Can Taste It." By Eliot Van Buskirk. 17JUN2009. Wired.com</span><br />
<br />
<span style="font-size: x-small;">"Winning Teams Join to Qualify for $1 Million Netflix Prize." By Eliot Van Buskirk 26JUN2009. Wired.com</span><br />
<br />
<span style="font-size: x-small;">"How the Netflix Prize Was Won." By Eliot Van Buskirk. 22SEP2009. Wired.com</span><br />
<br />
<span style="font-size: x-small;">"Netflix Prize: It Ain’t Over ’til It’s Over." By. Eliot Van Buskirk. 12AUG2009. Wired.com</span></span></div>CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.comtag:blogger.com,1999:blog-4323599490342956532.post-44338631019639119162011-05-01T22:41:00.001-04:002012-04-29T21:17:40.545-04:00Force Multipliers, Ballistics Software and the Tank Battle of 73 Easting<div dir="ltr" style="text-align: left;" trbidi="on">
<div style="border-bottom: medium none; border-left: medium none; border-right: medium none; border-top: medium none;">
<br />
<div class="separator" style="clear: both; text-align: center;">
</div>
I heard stories for years about a near mythical Desert Storm tank battle, where every Iraqi tank was destroyed while all of the vastly outnumbered American tanks survived. A long search revealed this as the Battle of 73 Easting <a href="http://www.youtube.com/watch?v=aBG_G678Trg">(3 minute documentary clip here)</a>. Eight M1A1 'Abrams' and 12 Bradleys destroyed<em> "28 tanks, 16 personnel carriers and 39 trucks in 23 minutes. With no American losses."</em><br />
<em></em></div>
The story continues though, because the American Cavalry unit was attacked for the next 6 hours, calling in help from attack helicopters and hundreds of artillery barrages as they lost one tank and slowly ran out of ammunition (having run out of anti-tank missiles early in the engagement). At some point reinforcements arrived for the Americans, trebling their total force so I am unsure how many of the <em>"</em><a href="http://en.wikipedia.org/wiki/Battle_of_73_Easting"><em>160 tanks, 180 personnel carriers, 12 artillery pieces</em></a><em> and more than 80 wheeled vehicles" </em>destroyed are attributed to the original <em> </em><a href="http://www.blogger.com/"></a>20 American tanks.<br />
<br />
Such a remarkable story begs the question, How did they destroy so many tanks without losing any themselves?<br />
<br />
<h1>
Documented Advantages</h1>
<br />
Most articles talk about the importance of surprise in the Battle of 73 Easting, but other sites emphasize that tanks are <u>loud</u>, so they rarely surprise enemies. Other articles mention the M1A1's superior <a href="http://www.fprado.com/armorsite/abrams.htm">firing range</a> over enemies, but the Wikipedia entry mentions the Americans <a href="http://en.wikipedia.org/wiki/Battle_of_73_Easting">"cresting a low rise and surprising an Iraqi tank company"</a> thereby implying that some engagements were in close quarters. Others cite American tanks' rapid refire rate of 3 seconds per shot as the reason, but the engagement lasted <em>hours</em>. The book <u>Warrior's Rage</u> explains that a sandstorm preceded the engagement, so attack helicopters were not the cause. Some articles cite the superior imaging and targeting capabilities of American tanks, but this equipment only existed in M1A<strong><em>2</em></strong> tanks which were not in the fight. If none of these capabilities were the reason, then how did they destroy ten times as many of the enemy?<br />
<br />
<h1>
Fire Control</h1>
<br />
When I heard the 'myth' of this battle, I was told that American tanks' ability to fire while moving was the critical advantage. The Iraqi tanks needed to stop, aim and fire to destroy their opponents. I can't find a single article on the internet that says U.S. tanks can fire while moving, but I have found lots of indications this is true.<br />
<br />
<div style="border-bottom: medium none; border-left: medium none; border-right: medium none; border-top: medium none;">
<object class="BLOGGER-youtube-video" classid="clsid:D27CDB6E-AE6D-11cf-96B8-444553540000" codebase="http://download.macromedia.com/pub/shockwave/cabs/flash/swflash.cab#version=6,0,40,0" data-thumbnail-src="http://3.gvt0.com/vi/gzfLvzn2_Sc/0.jpg" height="266" width="320"><param name="movie" value="http://www.youtube.com/v/gzfLvzn2_Sc&fs=1&source=uds" />
<param name="bgcolor" value="#FFFFFF" />
<embed width="320" height="266" src="http://www.youtube.com/v/gzfLvzn2_Sc&fs=1&source=uds" type="application/x-shockwave-flash"></embed></object>Tank personnel call the targeting software 'fire control'. On YouTube I found a <a href="http://www.youtube.com/watch?v=gzfLvzn2_Sc">video demonstration</a> of the M1A1 Abrams tank fire control simulator. Interestingly, it shows many simulations of firing while moving. Furthermore, there is a quote from <u>Warrior's Rage</u> (written by the Major in command of the 73 Easting force) which says, <em>"The idea of diggin in tanks and armored fighting vehicles designed to fire on the move and smash through enemy defenses at 35 or 40 mph made no sense to me."</em> <br />
<br />
If this evidence of "fire on the move" capabilities wasn't enough, the following excerpt mentions that <em>"<a href="http://www.army-guide.com/eng/product.php?prodID=429">The fire control computer</a> automatically calculates the fire control solution based on: lead angle measurement; bend of the gun measured by the muzzle reference system; <strong>velocity measurement </strong>from a wind sensor on the roof of the turret; data from a pendulum static cant sensor located at the centre of the turret roof. The operator manually inputs data on ammunition type, temperature, and barometric pressure."</em> The reference to velocity measurement clearly indicates that the tanks' speed and direction are factored into firing solutions.<br />
<br />
This last excerpt clearly states that computers make these adjustments, but at its core though, this capability is a combination of elaborate and complex physics equations... and therefore is a Competitive Advantage via Quantitative Methods.</div>
<br />
<h1>
Moving Targets Are Hard to Hit</h1>
<br />
If you're moving at 40 mph and can hit a stationary enemy tank which can't return fire becase you're moving, you'll win every time. To fully appreciate the complexity of this capability, I would ask, "Can you sink a basketball while you're jumping left to right?" Do you think you can do it while you <em>and</em> the basket are moving? Do you think you can do it when you and the basket are moving in opposite directions? Well, that's what this fire control software does.<br />
<br />
<h1>
Benefits and Value Quantification</h1>
<br />
<a href="http://www.google.com/#hl=en&q=Force+multiplier&tbs=dfn:1&tbo=u&sa=X&ei=Y8C8TaaCOoHagAeDlqTgBg&sqi=2&ved=0CBkQkQ4&bav=on.2,or.r_gc.r_pw.&fp=40d12f84dda1c17c"><em>Force multiplication,</em></a><em> in military usage, refers to an attribute or a combination of attributes which make a given force more effective than that same force would be without it. The expected size increase required to have the same effectiveness without that advantage is the multiplication factor.</em> So <strong>a force multiplier is a competitive advantage</strong>.<br />
<ul>
<li><strong>10 to 1 kill ratio</strong></li>
<li><strong>Kill ratio is even higher:</strong> Virtually zero American losses, which indicates the kill ratio would be higher if the enemy had more tanks.</li>
<li><strong>Game Theory: Firing Speed Matters. A LOT.</strong> 1 tank that fires 10x faster than the enemy is worth 10 enemy tanks (if it survives the first barrage). This includes aiming time as well as loading time.</li>
<li><strong>Game Theory: Accuracy and Lethality Matter:</strong> For every round that hits home, you have one fewer enemy shooting back at you. So any investment that increases firing accuracy, or the lethality of each round is well worth it. This is probably why the U.S. Army uses depleted Uranium shells while other countries use tungsten. </li>
<li><strong>Greater survivability = greater return on existing capital: </strong>Allows you to exhaust your opponent financially (strategy a la <a href="http://en.wikipedia.org/wiki/Charlie_Wilson's_War">Charlie Wilson's War</a>).</li>
<li><strong>Psychological Warfare Benefit</strong> - The Iraqi Tank Divisions didn't bother to fight in 2003. They just <a href="http://www.freerepublic.com/focus/f-news/872245/posts">surrendered</a>.</li>
<li><strong>Dollar Efficiency:</strong> After you've invested $5 million in metal, treads, engines, etc, why not spend a few extra bucks on software that makes you invincible?</li>
</ul>
<h1>
Commentary</h1>
<br />
What is most interesting is that the fire control (aka targeting) software that creates these competitive advantages is not cited as important anywhere in the media. Is the U.S. not emphasizing the 'fire on the move' capability because then enemies would use countermeasures (stay moving, rapid changes in direction, pinning American tanks before firing, always traveling at top speed when under attack and never traveling in straight lines)? <br />
<br />
To change topics and talk about U.S. military spending, perhaps the U.S. doesn't need anymore tanks? I'd support the development of more sophisticated fire control software, but increasing quantity just seems like overkill at this point.<br />
<br />
<br />
<a href="http://en.wikipedia.org/wiki/Battle_of_73_Easting">http://en.wikipedia.org/wiki/Battle_of_73_Easting</a><br />
<br />
<u>Warrior's Rage.</u> By Douglas MacGregor. Copyright 2009. Naval Institute Press, Maryland.</div>CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.comtag:blogger.com,1999:blog-4323599490342956532.post-17464413589301526932011-03-13T22:21:00.000-04:002011-12-17T12:20:53.386-05:00Text Mining, Legal Research and Google Books<div class="separator" style="clear: both; text-align: center;"></div><div class="separator" style="clear: both; text-align: center;"></div><div class="separator" style="clear: both; text-align: center;"></div><div style="border-bottom: medium none; border-left: medium none; border-right: medium none; border-top: medium none;">
<h1>What does text have to do with quantitative methods?</h1>
</div><br />
<div style="border-bottom: medium none; border-left: medium none; border-right: medium none; border-top: medium none;">Text mining is the use of computers and algorithms to classify, associate, or search large amounts of <u>unstructured</u> text data. This is a quantitative method because it involves pattern recognition… a traditionally quantitative subject. Further, the website of one text mining company,
<a href="http://www.autonomy.com/">Autonomy</a>, states that their software was <em>“built on the seminal works of Thomas Bayes and Claude Shannon…”</em> the fathers of <a href="http://en.wikipedia.org/wiki/Bayesian_inference">Bayesian statistics</a> and <a href="http://en.wikipedia.org/wiki/Information_theory">information theory</a>, respectively.</div><div style="border-bottom: medium none; border-left: medium none; border-right: medium none; border-top: medium none;"><br />
</div>
<div style="border-bottom: medium none; border-left: medium none; border-right: medium none; border-top: medium none;">
<h1>TF-IDF and PageRank</h1>
</div>
<div style="border-bottom: medium none; border-left: medium none; border-right: medium none; border-top: medium none;">To further demonstrate the quantitative nature of text mining, I present the <em>‘Transaction Frequency – Inverse Document Frequency’</em> (<a href="http://en.wikipedia.org/wiki/Tf%E2%80%93idf">TF-IDF</a>) metric. Documents can be grouped according to the similarity of their contents using this single metric… based strictly on word counts and the count of document appearances for a word. If a word appears in every document (such as ‘and’ ‘or’ ‘the’ etc) this metric ignores that word. If a word appears in only two documents and that word appears often (such as ‘mince’ in cookbooks, ‘precedent’ in legal texts, or ‘love’ in romance novels), then this metric says those two documents are very similar. </div><div style="text-align: center;"><span style="font-family: "Times New Roman"; font-size: 12pt; mso-ansi-language: EN-US; mso-bidi-language: AR-SA; mso-fareast-font-family: "Times New Roman"; mso-fareast-language: EN-US;"><shapetype coordsize="21600,21600" filled="f" id="_x0000_t75" o:preferrelative="t" o:spt="75" path="m@4@5l@4@11@9@11@9@5xe" stroked="f"></shapetype></span><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgFlPNq__qehZouubZjZvzJaNz4YE6t6zJeYBb6arPN6bjU1Ui5oQygzsvkAz22SwdOBdKZaWshrl87w-qno98fg73gB-uQ_rOaJ6J5tjGyWW8Ku9uqObtw8pSjagKObEruDh_9r6TjPzPo/s1600/tfidf.bmp" imageanchor="1" style="margin-left: 1em; margin-right: 1em;">
<img border="0" height="40" q6="true" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgFlPNq__qehZouubZjZvzJaNz4YE6t6zJeYBb6arPN6bjU1Ui5oQygzsvkAz22SwdOBdKZaWshrl87w-qno98fg73gB-uQ_rOaJ6J5tjGyWW8Ku9uqObtw8pSjagKObEruDh_9r6TjPzPo/s320/tfidf.bmp" width="320" /></a></div><div style="border-bottom: medium none; border-left: medium none; border-right: medium none; border-top: medium none;">If you still don’t believe that text mining is quantitative, then check out the Wikipedia entry for
<a href="http://en.wikipedia.org/wiki/PageRank">PageRank</a> (the Google Search algorithm). Google prioritizes websites based on the words you pick, the words on each website, and where the sites are linked, but it is using a great deal of math to do so.</div><br />
<div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEg86cPd9Eyc3gMd_0-Kg51SuV6Ciz0GSRbI4vSyORjDnyEvcy7KheHY7lth5QFG3czrEhkWqacAmpSIAIJ2WtrHGVpeJxRDX01qXdFKI_UCfcA3hbHTBq1ArGpYmOo1fQFJfCo5cjANp6JP/s1600/pagerank.bmp" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" height="213" q6="true" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEg86cPd9Eyc3gMd_0-Kg51SuV6Ciz0GSRbI4vSyORjDnyEvcy7KheHY7lth5QFG3czrEhkWqacAmpSIAIJ2WtrHGVpeJxRDX01qXdFKI_UCfcA3hbHTBq1ArGpYmOo1fQFJfCo5cjANp6JP/s320/pagerank.bmp" width="320" /></a></div>
<h1>Existing Applications</h1><br />
The most useful specializations of text mining concern information extraction (How can I determine the type of crime from a police report) and information retrieval (Which documents in the library are relevant?) These technologies are revolutionizing knowledge work, as <a href="http://books.google.com/">Google Books</a> makes it possible to search millions of documents for specific phrases and word combinations, <a href="http://mail.google.com/support/bin/answer.py?answer=6603">Gmail </a>uses email text to target advertising, and the FBI’s
<a href="http://en.wikipedia.org/wiki/Carnivore_(software)">Carnivore</a> program scans every email in the United States for indications of criminal or terrorist activity.
<a href="http://www.nytimes.com/2011/03/05/science/05legal.html"><em>“We’re at the beginning of a 10-year period where we’re going to transition from computers that can’t understand language to a point where computers can understand quite a bit about language.”</em>
</a><em>
</em><br />
<br />
<h1>A Few Other Existing Applications</h1><br />
• <strong>ClearForest</strong> – Software that combs through financial newsfeeds to detect news that will affect company prices. Wall Street firms can then write software that automatically sells insurance firm stocks if an earthquake occurs in their coverage zone for example.<br />
• <strong>Teneros Social Sentry</strong> – Large companies purchase this service which scans Twitter, Facebook, LinkedIn, mySpace, Orkut, and blog postings of all employees for criticism of the employer, its customers, inappropriate behavior, prospecting, and data loss prevention.<br />
• <strong>Email Filtering</strong> – Your inbox would be full of spam if companies couldn’t filter it out based on textual analysis.<br />
<br />
<strong>Benefits and Value Quantification</strong><br />
• <strong>Scalable</strong> – Information overload is no longer a limitation.<br />
• <strong>Breadth</strong> – You can search every document every written.<br />
• <strong>Uniformity of Approach</strong> – Whereas two legal researchers’ criteria may differ, an algorithm is applied uniformly for all instances.<br />
• <strong>Accuracy</strong> – Neither fatigue, distraction, nor habituation will impair an algorithm’s accuracy. <br />
• <strong>Unbiased</strong> – Subjectivity isn’t a problem.<br />
•<strong> Speed</strong> – <em>“</em><a href="http://www.clearwellsystems.com/"><em>Clearwell’s</em></a><em> software was used… to search through a half-million documents… [it] analyzed and sorted 570,000 documents… in two days. [It] used just one more day to identify 3,070 documents that were relevant to the court-ordered discovery motion.”</em><br />
• <strong>Cost </strong>– <em>“In 1978, six television networks paid $2.2 million dollars ($7,752,096 in 2011 dollars) in legal fees to review 6 million documents."</em> This now costs less than $400,000… <u>a 95% price reduction through text mining algorithms and software.</u><br />
• Focus on <strong>analysis, not data collection</strong> – The founder of Autonomy <em>“estimated that the shift from manual document discovery to e-discovery"</em> would allow one lawyer to do the work previously done by 500, and that the next generation of software will again double productivity.<br />
<br />
<h1>Commentary</h1><br />
If useful information is an iron needle and information is the haystack, then Google, Autonomy, Lexis Nexis and their competitors are very powerful electro-magnets. What opportunities are created by the ability to find the needle of information in the haystack? Will library sciences exist a decade from now? <br />
<br />
If knowledge is power, how valuable is technology that fights information overload?<br />
<br />
What happens in a world where knowledge is instantly accessible from any source? Speeches, webpages, every book every written… and it is already prioritized based on your search? With information extraction essentially free, the future will go to those who are the most creative at integrating that information (more to follow on this topic in my Connotate.com blog entry). <br />
<br />
<br />
“Armies of Expensive Lawyers, Replaced by Cheaper Software.” By John Markoff. New York Times. March 5, 2011 <br />
<br />
“Falling Demand for Brains?” By Paul Krugman. New York Times. March 5, 2011.<br />
<br />
<u>Text Mining. Predictive Method for Analyzing Unstructured Information.</u> By Sholom Weiss, Nitin Indurkhya, Tong Zhang, and Fred Damerau. 2005 Springer Science+Business Media Inc. Pg 85, 178-182CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.comtag:blogger.com,1999:blog-4323599490342956532.post-28668826145680665532011-03-04T21:19:00.000-05:002011-12-17T12:18:06.591-05:00Capital One and Microsegmentation: “Your business magic lies in the algorithms of customization…”<h1>Credit Cards… Late 20th Century</h1><br />
The credit card industry in the eighties was defined by uniform, non-tailored interest rates, imprecise credit risk decisions based on human intuition, little product development, and universally adopted metrics of cardholder risk. Credit cards were also highly profitable, although credit losses varied widely by issuing bank.<br />
<br />
Capital One achieved phenomenal success in this industry by adopting their <em>Information Based Strategy</em> at a time when no rivals cared to change. More specifically, '<em>What we want are markets where people can see the product being sold, but not the algorithms behind it. Your business magic lies in the algorithms of customization…'</em> They then pursued this strategy through the application of quantitative methods to create a powerful competitive advantage: scalable micro-segmentation.<br />
<br />
<h1>Scalable Microsegmentation</h1><br />
Capital One’s CEO summarized their strategy as, <em>‘The right product to the right customer at the right time for the right price.’</em> This required the replacement of manual credit review because the such processes <em>('relying on your tummy to make [credit] decisions')</em> were neither scalable nor did they allow new insights to be readily transferred to other analysts. <em>'[CapOne management also] claimed that their complex statistical models would be <u>better</u> than credit analysts at distinguishing between creditworthy cardholders and their apparently similar, but unworthy peers.'</em> The use of these statistical models to replace human intuition made these processes <strong>instantaneous</strong>, <strong>less labor intensive</strong>, <strong>unbiased</strong>, and for the first time <strong>scalable</strong>. Once such risk models were implemented, their results could be used as an input in pricing decisions. The interest rate necessary to cover defaults and operational costs would be set as the absolute minimum, while the profit margins could be adjusted based on the expected longevity of the customer relationship, projected changes in spending behavior or repayment patterns.<br />
<br />
<h1>Applications of Quantitative Methods</h1><br />
• <strong>Credit Risk: </strong>Loan officers were replaced by <em>‘complex statistical models’</em> of risk, thereby eliminating human bias.<br />
• <strong>Customized Pricing: </strong>Stopped offering $20 annual fees and 19% interest rates to all customers, and used on statistical models to maximize profitability and customer acquisition.<br />
• <strong>Value Determination:</strong> Lifetime NPV is calculated for every cardholder.<br />
• <strong>Customer Retention: </strong>If the expected customer NPV turns negative, CapOne stops retention efforts.<br />
• <strong>Product Life-cycling:</strong> When the customer lifetime NPV drops from one campaign to the next (as competition sets in) they drop the product from their line. (This is more of an <em>'actuarial approach'</em> to risk.)<br />
• <strong>Sales Prediction: </strong>CapOne knows which incremental products they want to sell their existing customers at all times, even tailoring the choices to the channel by which they’re contacted. <a href="http://news.slashdot.org/story/10/11/04/132257/Do-Firefox-Users-Pay-More-For-Car-Loans">(i.e. Whether you access their website with Chrome, Firefox, or Internet Explorer determines the interest rates you’re offered)</a><br />
• <strong>Sales & Customer Service Optimization: </strong><em>‘In the time that phones took to generate their first ring, Capital One’s computers identified the caller and predicted – with greater than 70% accuracy – what the caller wanted, and what the caller might be willing to buy.’</em><br />
• <strong>Best Practice Propogation:</strong> Sharing best practices was simplified because they were implemented through quantitative methods rather than through intuitive, human intensive systems. <br />
<br />
• <strong>Information Extraction:</strong> CapOne held approximately 5 single spaced pages of information on every person in the United States, and used this information to make their pricing, underwriting, and acquisition efforts more powerful. Deriving value from this much information is only possible with data mining methods. <br />
<br />
<h1>Performance Analysis</h1><br />
• <em>“From its IPO in 1994 to 2000, CapOne’s stock price had increased more than 1,000% while the S&P500 had increased just under 300%.”</em><br />
• <em>“[CapOne’s Parent company] was the best performer on the NYSE... until it spun off CapOne”</em> in 1994.<br />
• <em>“The company’s average annual growth rate of 46% was the highest in the industry (excluding growth through mergers and acquisitions).”</em><br />
• Grew from <em>“a minor credit card subsidiary of a small, regional bank – and turned it into the nation’s seventh largest credit card issuer.”</em><br />
• $1.1 billion IPO in October 1994 when the credit card division was spun off into an independent company.<br />
• Credit card revenues grew from 25% of company revenues to 67% from 1988 to 1994.<br />
<br />
<h1>Personal Thoughts</h1><br />
I first read this case as an undergrad, and it made a strong impression on my professional future. Rereading the case though, it highlighted two benefits of quantitative methods that I had forgotten: Transferability and Scalability. Statistical models are highly portable… a new employee can see the model and understand the company's underwriting standards quickly. Furthermore, CapOne didn’t need to increase the size of their underwriting department by 46% annually despite revenues growing that much because the decision process was automated. This stands in sharp contrast to <a href="http://www.straightfromthegut.com/">GE, where they quickly hired 3,000 loan officers upon entering the Thai auto loan market.</a>.. no wonder that CapOne’s operation was growing faster, even though GE was competing in the credit card industry at the same time! <br />
<br />
The case also used an industry buzz word that I previously didn’t associate with quantitative methods: Micro-segmentation. <strong>Micro-segmentation requires customized pricing, customer service, risk evaluation and you can’t achieve this without quantitative methods. </strong>Some will say that companies don’t need quantitative methods for this… that they could just classify 100 market segments. I’d argue though that treating customers as individuals requires that pricing and risk cover the continuum of possibilities. <br />
Lastly, CapOne’s CEO also emphasized a profound advantage of quantitative methods by stating their pursuit of, <em>'Markets where the algorithms behind the product can’t be reverse engineered.'</em> Quantitative methods are therefore valuable <strong>because of the obfuscation</strong> built into them. It is time consuming, resource intensive, and difficult to reverse-engineer algorithm based business processes and that difficulty makes them more valuable competitive advantages.<br />
<br />
‘Capital One Financial Corp’ Harvard Business School Case 9-700-124 Revised May 1, 2001.<br />
<a href="http://cb.hbsp.harvard.edu/cb/product/700124-PDF-ENG">http://cb.hbsp.harvard.edu/cb/product/700124-PDF-ENG</a><br />
<br />
<u>Straight from the Gut. </u>Welch, Jack. Copyright 2001. Warner Books, New York, NY.CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.comtag:blogger.com,1999:blog-4323599490342956532.post-68657160328685566132011-02-21T17:44:00.001-05:002011-12-17T12:17:14.996-05:00Malthus, Demographics, Scrooge and Rwanda<div dir="ltr" style="text-align: left;" trbidi="on">Kevin Spacey in the T.V. show <em>Wiseguy</em> says, <br />
<br />
<div></div><em><span style="font-size: large;">"The population grows geometrically and the food supply grows arithmetically. Three things keep the balance: Famine. Disease. And war." </span></em><br />
<br />
<div></div>This is an accurate synopsis of Thomas Malthus' <a href="http://www.gutenberg.org/cache/epub/4239/pg4239.txt">An Essay On The Principle of Population</a>, published in 1798. Although Malthus never words this so darkly, he does say, <em>"The power of population is indefinitely greater than the power in the earth to produce subsistence for man". </em>This is because populations grow at a constant rate (e.g. 5% per year) while farm output tends to grow at a fixed amount per year (e.g. 5 additional pounds of corn per acre). Eventually, the compounding of the population outstrips the food supply, or as Malthus put it <a href="http://en.wikipedia.org/wiki/Thomas_Robert_Malthus">"the increase of population is necessarily limited by the means of subsistence."</a><br />
<br />
<br />
<div>
<h1>Empirical Proof</h1>
</div>As evidence that the population grows faster than food production, Malthus cites the aftermath of pandemics and famines around the world, <em>"The effects of the dreadful plague in London in 1666 were not perceptible fifteen or twenty years afterwards. The traces of the most destructive famines in China and Indostan are by all accounts very soon </em><em>obliterated. It may even be doubted whether Turkey and Egypt are upon an average much less populous for the plagues that periodically lay them waste. "</em><br />
Much more troubling though, is the idea that Malthusian economics continues to exist today... <br />
<br />
<div>
</div>
<h1>"If they would rather die they had better do it, and decrease the surplus population." - Ebenezer Scrooge</h1>
<br />
The book <u>Collapse</u> by Jared Diamond (of <u><a href="http://en.wikipedia.org/wiki/Guns,_Germs,_and_Steel">Guns, Germs, Steel</a></u> fame) analyzes several accounts from Rwanda... many pointing to Malthus' prescience. More specifically, <u>Collapse</u> emphasizes that while the genocide of Tutsis was ethnically motivated, murders still occurred in villages <em>that were entirely Hutu. </em>The book also does the straightforward arithmetic of:<br />
<br />
<div style="text-align: center;">
<strong>Acres of Arable Rwandan Land / Population = 1/2 Acre per Person</strong></div><br />
<div>It is impossible to feed someone with the agricultural output from a half acre of farmland, but the book further highlights the underlying cause of the war by quoting Rwandans saying, <em>"We need another war. We still have too many people."</em>
</div><br />
<div></div>
<h1>Commentary</h1><br />
I think anyone will agree that the mathematical argument is valid, but the assumptions underlying the math are suspect.<br />
<ul>
<li>Population grows faster than food production.</li>
<li>Food production grows by a constant <em>amount</em>, but can't grow at a constant <em>rate</em>.</li>
<li>Population grows at a constant rate.</li>
</ul><br />
<br />
<div>There is ample evidence of populations growing at different rates (for example Europe isn't growing while North Africa is), but remember that contraceptives didn't exist in Malthus' day... so populations couldn't slow their growth rates down. Even when a population stabilizes around a level of sufficiency, it can overshoot due to unforeseen flooding, frost or drought, resulting in famine. Furthermore, agricultural productivity may be increasing, but will food output per acre always increase? Western countries may have the luxury of outsourcing food production to other countries, but developing countries can't... especially those in the grip of Malthusian economics. </div>
<div>
</div>
<div>I consider it especially profound though, that on the basis of a quantitative argument, a demographer was able to predict wars at a time when society believed that the Industrial Revolution and modern science would eliminate famine and war forever. As Malthus demonstrated, eliminating famine is difficult with an increasing number of mouths to feed, and impossible when they're increasing faster than the food supply.
</div> <a href="http://www.gutenberg.org/cache/epub/4239/pg4239.txt">"An Essay on the Principle of Population"</a>, 1798 by Thomas Malthus.</div>CAvQMhttp://www.blogger.com/profile/07078494183679441527noreply@blogger.com