SECTOR / POLICY
Test the message before it goes public.
See how people respond to each message. Get the effect and the uncertainty.
0-7 days
Typical benchmark polling fieldwork takes most of a week. Simulated results do not replace polling or represent observed voter behavior.
THE DECISION
Which message changes support, and by how much?
Compare estimated responses to election and issue messages. Each result shows the effect and its uncertainty. We do not test persuasion that deceives. Our acceptable use policy bans it.
VALIDATED WEEKLY
0.73
Mean rank agreement
43 design-filtered human studies
Confidence, with every report.
Every report carries graded confidence. Each week, we test the system against held-out human data and publish the full benchmark, including mid-range results and failures.
Read the validation paper ↗