Small Data Beat Big Data In Election 2012

Social Studies: Obama vs. Romney

Even before polls closed Tuesday, some observers were describing the Democrats' vaunted "ground game" -- a.k.a., get-out-the-vote, or GOTV -- as a victory for big data.

Doubtless, the staff of both presidential candidates performed some deep analysis of large datasets. Having participated with my family in GOTV efforts for the past four quadrennial U.S. elections, I have a different take: The Obama re-election machine was really about small data.

Small Data Writ Large

Compared to previous years, what was fascinating about the Obama effort this time was the intense effort on data curation at the individual record level. In past cycles, registration and voter data seemed like a read-only Excel sheet: You got a well-thumbed, printed list of voters to call or visit with some basic info about them and -- if you were lucky -- some handwritten notes from previous canvassers. This led to a lot of duplicate effort, frustration and miscommunication.

This time, we saw a much more systematic effort to refine and annotate each record iteratively, to match message and approach to where someone lay in their decision cycle. When someone was "touched," their record got updated. If the previous canvasser uncovered a preference for Obama, the next canvasser worked on intent to vote. And if that succeeded, the next canvasser, closer to Election Day, worked to make sure the voter had a practical plan for voting (a key predictor of actual voting, especially among young or distracted voters). A premium seemed to be placed on obtaining mobile phone numbers, updated address information and names of additional potential voters in the household (typically adult children living at home).

[ For more on how the campaigns used big data to chase voters, see Election 2012: Who's Winning Big Data Race? ]

An enterprise data manager would recognize the basic steps: