Full transparency on our methodology. We show our work so you can decide if our grading is fair.
We grade each representative on five dimensions of accountability. Each component is weighted based on how important it is to accountability:
Do their votes match their stated positions? Do they flip-flop? Do they align with their constituents' interests?
What we measure: Position contradictions (major = -10 points, moderate = -5 points, minor = -2 points), voting flip-flops over time, constituent alignment based on donor geography and district demographics.
Do they use language that meets a nonpartisan standard of civil discourse? Or do they rely on dehumanizing rhetoric, incitement, and personal attacks?
What we measure: Keyword scanning across all posts and transcripts. Dehumanizing terms and incitement language carry 3× weight; personal attacks carry 1×; common partisan rhetoric carries 0.3×. The same words trigger the same penalty regardless of party.
How transparent are they about their positions and reasoning? Do they engage with media across the political spectrum or only appear on friendly networks?
What we measure: Number of stated positions on 20 key policy topics, media appearances weighted by substance quality and ideological cross-pollination. One substantive interview on opposing-lean media counts more than 15 friendly talking-points hits.
Do their public statements contradict each other? Do they say one thing and then say the opposite?
What we measure: Rate of self-contradictions in social media posts and public statements per 1,000 posts. Rate-based, not count-based, so active communicators aren’t unfairly penalized.
Are they focused on their district or performing for national cameras?
What we measure: Donor geography alignment (in-district vs. national funding), local social media engagement, town hall frequency and accessibility.
We grade consistency, not ideology. A conservative who votes conservatively and a liberal who votes liberally both get high grades (as long as they're consistent). A conservative who votes one way and talks another gets a low grade, regardless of party.
Contradictions must be genuine. We filter out false positives with five guardrail rules: the positions must be at least 30 characters (to rule out frivolous claims), directly related to the vote, not omnibus bills, not procedural tricks, and pass a "is this hypocritical?" test with AI review.
We show our work. Every contradiction includes the source (link to the vote, quote, date). You can verify our work yourself. We encourage it.
Data is updated daily. Grades are recalculated as new votes happen. Last vote was yesterday. Next update is tomorrow. You're always looking at current information.