Transparency
StateThat shows civic sentiment for bills and current event topics. The signal quality layer helps us understand how reliable that sentiment is before we use it in embeds, share cards, or future public summaries.
We use trust-weighted civic sentiment to separate a simple vote count from a healthier read of community opinion. Vote totals still matter, but the quality layer also looks at whether a sample is large enough, whether verified participants are represented, and whether the group resembles the public jurisdiction being summarized.
Public bill and topic percentages remain the same. Signal quality metadata is an attribution layer that helps readers understand when a snapshot was calculated and whether it came from live votes, legacy-derived backfill, or a mix of sources.
Raw sample size is the number of votes included in a snapshot. The effective sample size is a more conservative estimate that accounts for weighting, because a heavily weighted sample can carry less statistical information than its raw count suggests.
ACS fit compares available demographic distributions against Census American Community Survey benchmarks for supported jurisdictions and districts. The goal is not to identify people. It is to flag when a sample is unusually concentrated by age, race or ethnicity, education, income, geography, or source.
Share of voters aged 18–29
Flagged: the sample skews young.
Some early snapshots may be legacy-derived. That means they were backfilled from votes that already existed before the audit layer captured source telemetry. These snapshots can help internal quality review, but they have limitations: they do not invent missing IP, device, user-agent, referrer, or campaign metadata.
When a public embed says Legacy-derived, Mixed source, Live signal, or Source unknown, it is describing the origin of the signal snapshot only. It is not a public confidence badge and it does not change the visible sentiment percentages.
The quality worker looks for anomalies such as vote velocity spikes, new-account surges, unverified surges, single-source concentration, and location instability. These checks help StateThat decide when a snapshot needs more review before it is used in broader distribution.
The 2–4pm burst is flagged as a vote velocity spike and held for review.
Every snapshot is checked for
Raw IPs, device IDs, user-agent strings, and trust scores are not public. Individual audit events are not public either. Public signal metadata is limited to safe snapshot fields such as the entity, calculation date, sample sizes, source counts, legacy share, methodology version, and already-public sentiment values.
Public
Never public
Every verified vote sharpens your district's reading.