Murmuration provides dozens of Model Scores to help you identify the right individuals for your organizing efforts. Within the Organizer platform, our Model Scores are available as attributes that you can use when creating a Community for an Outreach effort.
Our Model Scores provide predictions about the likelihood of individuals to have particular opinions, take specific actions, or have certain characteristics. For example:
- The Presidential General Turnout Score predicts an individual's likelihood of voting in the next Presidential election.
- The Partisan Score indicates a person’s likelihood to vote for Democrats in federal races (e.g. US Senate/House races).
- The Down Ballot Completion Score predicts whether voters are likely to fully complete their ballot in high turnout elections.
As we add to and improve our Model Scores, we publish Model Scores Release Notes. Each Model Score we build is accompanied by a release notes article that provides guidance on how the score was made and how to use it.
Methodology
We build our Model Scores using voter data, consumer data, demographic data, polling data, and other available, relevant data to the given model. Most of our Model Scores are built using gradient boosted trees (i.e., the algorithms that result from decision trees) to generate a score for every registered voter in the Atlas by Murmuration dataset. The scores are typically expressed as numerical values between 0 and 100, with higher scores indicating a higher likelihood of the prediction being true for each respective individual. For all the Model Scores we build, we exclude a random sample of the survey data from the modeling process as a test set to use for validation.
We also curate a limited number of third-party model scores and make them available to our Partners. These scores are built with similar methodologies to our own models. Whenever possible, we validate them against our data to ensure the highest quality.
Guidance for Turnout Scores
Each turnout score predicts a voter's likelihood of voting in that specific type of election. When choosing a turnout score you should consider the highest office being contested on the ballot on a given election day and select a score that reflects that race. For example, if you are working on a 2026 municipal election that falls on the same day as a Governor or US Senate race, you would want to use a Midterm General Turnout Score. But if the municipal election were the highest office on the ballot that day, you would instead use the Municipal Turnout Score (or in any case where that score is not available, the Local Voter Score).
Guidance for the Early Vote Score