On POLLING…
Unreliable polls gaming donor narrative. Prediction markets most people can't even read. And a bipartisan backlash on data centers that blindsided governors who thought they had community support. Same failure, three headlines.
Every one of these was a structural read failure, not a bad-luck news cycle. A poll samples what people say when a stranger asks. A prediction market aggregates what people are willing to bet. Neither one reads the thing that actually determines an outcome — how trust, tension, and organizing capacity move through a place before anyone's watching for it.
This is the moment to make the case plainly: the tools built to read trust structurally — not survey it, not bet on it — are exactly what the next six months need. The midterms are the test case the news is already writing for you.
We're RFUTR and PathSight Predictive Science, commercialized for campaign and civic-engagement use through Cohera Systems. We read trust structurally, at census-tract precision — no surveys, no self-reported answers, zero personally identifiable information collected at any stage.
Finds the most structurally disconnected precincts in a field — tested and held in the Election Day Test, a sealed, out-of-sample prediction ahead of a real election.
Reads how trust and organizing capacity actually move through a jurisdiction, independent of what anyone reports about themselves.
The variable a poll can't see: physical and institutional access doesn't track with income the way pollster assumptions do. Confirmed across three cities — this shows who's structurally reachable, not just who happened to answer the phone.
Prioritizes spend where the structural read shows real room to move, instead of spreading evenly out of caution.
Not Another Poll. Not a Betting Market Either.
Nobody is asked anything. Nobody profits from the answer. It's a structural read of how trust actually moves — tested and held on exactly what it claims, honest about what it hasn't proven yet. That's a genuinely different category from the two things currently failing in public, in real time, on the front page.
What this doesn't do: call a race or predict a margin. Floor Identification — finding the field's most disconnected ground — is tested and held, 3 of 4 in its first out-of-sample check. Ranking every precinct precisely by degree of connection is not, and we don't claim it is. That's the honest boundary, and it's exactly why the claim we do make is defensible under the scrutiny this news cycle is inviting.
The work behind this isn't new — it's the same instrument that's been read, corrected, and stress-tested across every practice area we run. What's new is the moment. Two major, ongoing news stories are making the case for a structural read of trust better than any pitch deck could. The next six months are when that case gets heard, or it doesn't.
If your plan for the next six months still leans on polling and hope, this is the moment to change that — before the next headline is about your race, not someone else's.