Methodology
Every number on Quorly comes from a documented method. This page is the documentation.
How Public Pulse is collected
Public Pulse is a voluntary community survey. Members answer standing civic questions; results update continuously. It is a measure of the Quorly community — not a probability sample of the general public — and it is labeled that way everywhere it appears.
Any signed-in member may respond once per question. Duplicate and automated responses are filtered.
Responses from identity-verified members are broken out separately so you can compare the two populations.
District-level counts use the district derived at signup. The address itself is discarded — see the Privacy Center.
Every Pulse result ships with its sample size, date range and a methodology note. Individual responses are never published, sold, or visible to anyone — including representatives.
How polling averages are computed
The Polling Center average is a transparency-weighted rolling average of published public polls. Three adjustments are applied, and all three are displayed openly on every average:
A poll's weight decays exponentially with a ~14-day half-life. A 6-week-old poll contributes roughly 12% of a fresh one.
Weight scales with the square root of sample size and is capped — a 5,000-person poll is not five times a 1,000-person poll, because sampling error is not the dominant error source.
Each pollster's persistent lean relative to the consensus is estimated and shown on its scorecard. We adjust for it and publish the adjustment — we never silently drop a pollster for its lean.
Polls enter the average only if they disclose sponsor, field dates, sample size, population and question wording. Every poll page shows its transparency checklist, and you can always see exactly which polls are in or out of an average — and why.
How the Representation Gap is calculated
For a given issue, the Representation Gap compares constituent sentiment (the share of a representative's verified in-district Pulse respondents supporting a position) with the representative's recorded action (their roll-call vote, sponsorship, or public position on the corresponding measure).
Example: 83% of a district's respondents support a stock-trading ban and the member voted against the ban bill → gap of 83 points on that issue. Gaps are computed only from recorded, citable actions — never from inferred or predicted positions.
How feed ranking works
The feed is ranked by four factors — and deliberately not by predicted engagement. Content does not rank higher because it makes people angry.
- Your interests: Topics, follows and districts you chose
- Recency: Newer civic events surface first
- Evidence quality: Sourced posts outrank unsourced ones
- Viewpoint diversity: Deliberate mix across perspectives
There is no outrage multiplier, no pay-for-reach, and no engagement-time optimization. Sponsored content, when it exists, is labeled and excluded from organic ranking entirely.
How forecasting reputation works
Reputation is earned only one way: by making forecasts that turn out to be right. Three published statistics define every forecaster:
Brier-style score across resolved markets — heavily wrong confident forecasts cost more than mild ones.
When you say 70%, does it happen about 70% of the time? Measured across probability buckets.
Scores stabilize with sample size; small samples are labeled provisional on leaderboards.
Q Credits cannot be purchased, so a large balance and a high leaderboard rank can only reflect forecasting skill — never spending power. Reputation cannot be bought, gifted, or transferred.
Data sources
| Source | Type | Covers | Update cadence |
|---|---|---|---|
| Congress.gov | Official legislative record | Bill text, status, sponsors, committee actions | Hourly |
| House Clerk roll calls | Official vote record | Every recorded House floor vote, member by member | Within minutes of a vote |
| Senate roll call votes | Official vote record | Every recorded Senate floor vote, member by member | Within minutes of a vote |
| FEC | Official campaign-finance record | Contributions, expenditures, committee filings | Daily |
| Federal Register | Official rulemaking record | Proposed and final rules, agency notices | Daily |
| Public pollster releases | Third-party research | Published polls meeting transparency criteria | As published |
The honesty rule: if a feed is down or stale, the affected figures show "Data temporarily unavailable" with the last successful sync time. We never display stale data as current, interpolate through an outage, or fill gaps with estimates.
AI-use policy
- AI is used to summarize bills and long documents in plain language. Every AI summary is labeled, links to the full official text, and is spot-checked by human reviewers.
- AI assists moderation triage, but no account is suspended and no content is removed by an automated decision alone — a human makes every enforcement call.
- AI never writes persuasive political content on the platform, never generates posts posing as users, and never ranks content by predicted emotional reaction.
- AI models are never used to infer an individual user's ideology, religion, or voting intention for advertising, targeting, or sale — that data is not built, so it cannot leak.
- AI-generated media shared by users must carry the AI-generated media label (see Community Standards).
Questions about a specific number? Every chart on Quorly links back to this page and to its underlying source. If you find an error, the correction policy on the Neutrality page applies.