Case Study
SECURING PERSUADABLE VOTER SUPPORT WITH MACHINE LEARNING
CAMPAIGNS & ELECTIONS REED AWARD WINNER, 2022
Best Use of Online Targeting for a Mayoral Campaign, Sandy Stimpson for Mayor of Mobile, AL
Initial Polling
For Sandy Stimpson’s campaign, data collection began in May 2021, polling 2,000 registered voters in Mobile, AL. Our research uncovered two models to aid voter contact decisions with the Sandy Stimpson for Mayor campaign: a support model to measure each voter’s support or opposition to Mayor Stimpson, and a challenger model to measure support or opposition for opposing candidates. Using machine learning, we developed a data sample from the scores of the polled 2,000 voters and extrapolated it to the entire Mobile, AL voter file - calculating a support/opposition score for every registered voter in Mobile.
Persuasion
Rather than maintaining the same strategy throughout the entirety of the campaign, utilizing and updating multiple audience models allowed us to refine the registered supporters/opposition voters in Mobile, Alabama while responding to findings in polling, and deliver the most effective messaging to both audiences, (with digital ads serving 4,584,567 impressions across all platforms). By pushing heavier persuasion messaging to our opposition audience throughout the course of this digital campaign, our supporter audience grew from an initial 44% in support of the Mayor in May to 58% one week out from the election, ultimately delivering Sandy Stimpson a decisive win for his third term as Mayor of Mobile with 62.5% of the votes. This 62.5% was a significantly higher winning percentage than Mayor Stimpson’s previous victories with 53.5% in 2013 and 57.05% in 2017.
Final Steps
Once configured, these voter universes were deployed via social and programmatic digital advertising to ensure that our most persuadable audience was receiving heavier persuasion messaging than decided voters. To further specialize our approach, we created custom creative tracks to align with specific target audiences. By identifying issues specific to different demographics in the Mobile community, we were able to craft relevant content that articulated Mayor Stimpson’s plan for each issue. Users who were identified as supporters of Mayor Stimpson were targeted with all digital tracks throughout the entirety of the campaign, but gradually received less persuasion messaging than our persuadable audiences. To ensure Stimpson’s supporters turned out at the polls, we maintained this approach in the final weeks of the campaign and continued to target potential voters with persuasion messaging, while identified supporters received GOTV messaging across all social and paid platforms.
