Sunday, December 16, 2012

Narwhal, Dreamcatcher, Houdini, Optimizer, Call Tool and Blaster - The ‘MoneyBall Election’ of 2012

The 2012 Presidential election has been labeled the ‘Moneyball Election’for the transformative role that big data and analytics played in the Obama campaign’s victory. On the other hand though, Nate Silver of the New York Times 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.
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.
“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.” (Ferenstein)
So the question then becomes, how significant were the technological additions to the political arsenal, and did any of them create a competitive advantage?

Quantitative Methods
Many technologies were deployed by the Obama campaign, but only a portion of them were quantitative methods.
  • A/B Testing: Dozens of tests performed per day to deliver incremental improvements in messaging, fundraising and site engagement.

    • For the button, an A/B test 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.” (Wired)
    • “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.” (Wired)
    • “Many of the e-mails sent to supporters were just tests, with different subject lines, senders and messages.” (Time)
    • “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.” (Atlantic)
  • Dreamcatcher: Text Mining and “Microlistening” (Slate)
    • Call lists prioritized by persuadability, and bythe message channel
    • “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.” (Time)
    • Sounds like they were relying on audience descriptions similar to those produced by Blue Fin Labs to select the best advertising slots. “As a result, the campaign bought ads to air during unconventional programming, like'Sons of Anarchy','The Walking Dead'and'Don’t Trust the B—- in Apt. 23', skirting the traditional route of buying ads next to local news programming.” (Time)
    • “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.” (Slate)
  • Facebook Blaster and Twitter Blaster: Leveraged microtargeting and volunteer’s the Facebook connections to promote voter registration of likely Democrats.
    • “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.’” (Atlantic)
  • Optimizer: Let media buyers substitute cheaper audiences for more expensive ones while advertising to the same target audiences.
    • “…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.“ (Atlantic)
  • Houdini: A project to let voters self-identify as democrats, and ensures that they get to the polls in battleground states.
    • “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.”(Gallagher)

Non-Quantitative Methods (But Foundational to CAvQM)
  • Narwhal: 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.
    • “Allowed individual app scaling” (Gallagher)
    • “Share one common data store” (Gallagher)
    • Call lists reconciled with fundraising lists, field workers, social media and marketing databases. (Time)
    Dashboard: Let’s anyone log in to the Obama campaign and immediately be given tasks to accomplish based on real-time data. “A ‘virtual field office’” (Gallagher)
    • “Helped automate recruitment and outreach of volunteers.” (Gallagher)

  • Call Tool: “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.” (Gallagher)
  • Identity: This program gamified all digital volunteer efforts and let them compete against one another.
    • “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.” (Gallagher)

    Quantification of Benefits
    • Lower Cost Per Impression: “Obama's campaign's cost per ad was lower ($594) than the Romney campaign ($666) or any other major buyer in the campaign cycle.” (The Atlantic) That’s particularly significant when campaigns spend more than $500 million on television ads.
    • Fundraising Optimization: “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].” (Christian)
    • Lift in Response Rate: “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.” (Atlantic)
    • Efficacy of Social Media: “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.” (Time)

  • The Obama campaign ran “the most tweeted tweet” (The Atlantic)
  • The Obama campaign ran “the most popular Facebook post” (The Atlantic)
  • Obama won. Twice.

  • Personal Notes
    Political Analytics as Table Stakes: 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.

    Incumbency Advantage Increases: Incumbency advantage now translates into IT preparedness and the greater technological sophistication of their campaigns. “‘I don't think we would have been able to [build Narwhal] if we had to deal with the primaries,’ Reed said.” (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?

    Narwhal as Real-time Communications Advantage: 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 replaces communication. No communication was wasted describing what had been done, who had been called already, etc. This communication also enabled dynamic resource allocation. Prior campaigns only had daily updates, meanwhile Narwhal updated in real time.

    Look for opportunities to automate communication with APIs.

    Centralization of IT Development: If your IT systems are fragmented with massive duplication of effort, IT centralization will be a substantial improvement. “’ 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.” (Gallagher)

    Data Driven Decision Making: A prior blog defined DDDM as a competitive advantage and the Obama campaign leveraged it fully. “Assumptions were rarely left in place without numbers to back them up.” (Time) as evidenced by the heavy use of A/B testing, and others noted that the campaign “took their data driven strategy to the next level.” (Atlantic)

    Microlistening: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.

    Crowdsourcing Political Activism: 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.

    “Cambridge’s BlueFin Labs decodes social media chatter” by Neil Swidey. The Boston Globe. November 25, 2012.

    “Pundit forecasts all wrong, Silver Perfectly Right. Is Punditry dead?” Techcrunch. 07NOV12.

    “When the Nerds Go Marching In.” 16NOV12. The Atlantic.

    “Project DreamCatcher: How Cutting Edge Text Analytics Can Help the President.” By Sasha Issenberg. Slate. 13JAN2012.

    “Nate Silver’s Political Calculus.”

    “Built to win: Deep inside Obama’s campaign tech.” By Sean Gallagher. November 14, 2012.

    “Consensus is boring! Washington still won’t pay big bucks for predictions.” By Nick Arnett. November 8th, 2012. VentureBeat.

    “Inside the Secret World of Data Crunchers Who Helped Obama Win.” Time magazine. November 7, 2012.

    “Our MoneyBall Election.” By Adam Gopnik. The New Yorker. 06NOV12.

    “The A/B Test: Inside the Technology That’s Changing the Rules of Business. ” By Brian Christian. Wired magazine. April 25, 2012.

    “Obama Uses Houdini.” The Daily Kos. 02NOV2008.

    Obama Campaign Introduction to Dashboard video.