Targeting Democratic campaigns with Mesa

November 12, 2021
Your organization has decided to make a difference by helping get Democrats elected around the country. Amazing! There are hundreds of thousands of races across the country and deciding which races to invest in and when is a tricky question. You want to make sure your investment is going to make a difference.

We, the team at Deck, built Mesa just for you. Deck has helped over 1,200 campaigns and organizations like Congresswoman Pramila Jayapal and Swing Left talk to the right voters through our scores and monitoring tools. We’re dedicated to making these tools widely available and accessible through our partnership with the DNC. We created Mesa to help your organization prioritize and target races while working across multiple districts– all at once.

Mesa uses the power of machine learning technology to develop predictive forecasts of candidate support and the likelihood of winning. These dynamic scores update daily with information like campaign finance, media mentions, and voter and candidate traits. Mesa’s forecasts will help you target your programs and find the best races to invest in across the country.

In Virginia in 2021, Mesa’s average error was between 1.9 and 3.3 points with an average error of 2.2 points. Going into election night, our scores were by far the most pessimistic and unfortunately for Democrats this cycle, were correct. We predicted Democrats would hold 49 seats in the House of Delegates and after all the votes were counted, they held 48. Deck predicted voters would cast 3.2 million ballots and 3.3 million people voted in Virginia. Deck saw the results of 2021’s election coming and with this sort of granular information, your organization can too.


So how did we do it? Most scores and models in the Democratic ecosystem rely on survey responses generalized to a broader audience. While high quality survey models do exist, they are usually expensive, don’t update regularly, and are often based on questions about generic democratic support instead of being specific to a campaign, location, and election.

Most importantly though, these models are less and less effective further down the ballot, especially in states or races that aren’t getting national attention and funding. Mesa’s internal research found that while generic scores were around 80% accurate for Presidential and Congressional elections, for local and state legislative elections they are only around 65% effective.

Mesa is able to produce high quality models and scores all the way down the ballot. Our approach, called contextual inference,  instead captures real data on individuals’ past behaviors and the context around those decisions to anticipate what people in new contexts might do in the future. We use data on voter and candidate traits, media, campaign finance, and past election results to predict the outcomes of each election. You can find more in depth documentation of our models including data sources and validation here. Importantly though, these predictions are accurate all the way down the ballot and help your organization prioritize outreach across the country. Mesa campaigns used the scores and models to make decisions and run better outreach in their districts in 2020 and 2021 with great success, and now your organization can too.

Mesa scores are not just more accurate down the ballot; they are also better value, faster, and easy to use. Through our easy to use web app, your organization can leverage Mesa’s models to develop IE field targets, run mail and phone programs, or digital campaigns.

Because our models are based on real voter data, they update daily to provide you with the most up-to-date information as the state of your races changes. And Mesa also allows your organization to start investing in races immediately. If you decide in early October you want to send mail to 5 local races in a state, we can get you set up with predictions, targets, and your mailing universe that day. Mesa is also great value, we give you predictions as well as targets for $1,000 per month per state plus export-fees. If your organization wants to work nationally, 5 states or more is $5,000/per month.

We are dedicated to making Mesa easy to use and understand for people of all data backgrounds which is why we have a team of dedicated support and software staff to make sure you get the most out of the app.

Our past customers love Mesa. Swing Left used Mesa in 2021 in Virginia and helped 8 of their targeted races win. They used Mesa’s predictions to make sure they spent at the right time.

To learn more contact [email protected] or schedule a demo with us here!

More Like This

Loading...