Innovating Campaigns with Precision

How do you win elections? You persuade more people to vote for your candidate, and turn out more of your supporters. But how do you know what messages and tactics will persuade and turn out more voters?

Most organizations just go with their “hunch,” their best guess based on what they think worked before. Sometimes they do research. But focus groups, or simply asking respondents what they think, for example — doesn’t work. Why? People are terrible at introspection and self-prediction.

You can’t measure the effectiveness of a message by asking a voter if she thinks a message will change her preferences. She doesn’t know. We have to observe how a message impacts preference distributions in a treatment group compared to a control group.

We combine randomized-controlled trials (RCTs) and machine learning algorithms to identify winning messages and persuadable voters. Only an RCT can tell you what works and what doesn’t. Then we head into the field with ads online, in the mail, or the air to make sure those voters move the way we want them to.

We use unique online survey samples that are pre-matched to an augmented voter file, which contains over a thousand data points about each voter. We identify the precise impact that a policy position or message has on outcomes and identify which voters are moved most using a toolbox of artificial intelligence/machine-learning techniques and cross-validation to ensure predictive accuracy.

We train a machine learning algorithm using the data gathered in the clinical message trial on each of the research subjects: the algorithm learns to predict the message’s impact on every registered voter in the state or congressional district, using hundreds of features from an augmented voter-consumer file. We then used this trained algorithm to predict the messages’ impacts for every registered voter, resulting in a list of the most persuadable/moveable voters.

We’ve conducted over a dozen experiments across five states in the last five years, identifying messages that work, and the voters we know we can move.
‍
We helped Kevin Cramer win in North Dakota, Mike Braun win in Indiana, and Josh Hawley win in Missouri. We moved targeted voters by close to +10 points in the Michigan Senate race – a winnable race had the funds been available.

We develop creative messages on a range of topics, but test them to make sure they will work. Sometimes, we find they don’t, or that a race isn’t worth investing in. For instance, we ran randomized-controlled message trials and modeled voters in Pennsylvania in 2018, but found the messages weren’t moving enough voters and the race was too far gone.

Knowing what not to do is a critical part of making an impact. Knowing that an ad will work gives us the confidence to go all-in. That makes us efficient and effective with investor funds.

And as a result of our efforts, we’ve built a data and knowledge for key 2022 races. We will pursue the winning of 3 governorships, 8 Senate seats, and 15-30 House seats. The future is America first—RINOs be damned.