Frequently Asked Questions

Voice is a political news and ratings app. You read short, sourced summaries of political Moments - votes, statements, decisions, scandals, endorsements, policy fights, and other events - then rate the politicians involved. Those ratings create a live public signal showing how voters are reacting to politics as it happens.

The goal is simple: make political feedback faster, clearer, and harder for institutions to ignore. Voice also powers Wonk, a companion data product for people and organizations that need to understand political sentiment in more detail.

Not in the traditional sense. A poll tries to estimate what a defined population thinks at a specific point in time. Voice measures how participating users - and, where available, verified constituents - react to specific political moments over time.

Raw Voice ratings are not statistically representative approval polling unless Voice clearly labels a score as a weighted estimate and explains the methodology.

A Voice score is the average rating submitted for a specific politician, issue, Moment, geography, or time period. The label matters: a Moment score means how users rated a politician in response to that event; a politician score summarizes ratings across many Moments; and a district or city score reflects the selected geography.

Voice shows rating volume and signal strength so readers know whether they are seeing an early reaction or a more established pattern.

Politics usually gives voters only a few slow feedback channels: elections, calls, donations, protests, polls, and social media noise. Voice adds a lighter-weight channel: rate what politicians do, see how others are reacting, and help create a public record of constituent sentiment.

One rating is not a revolution. Many ratings, tied to real political geography and real political events, can't be ignored.

A politician does not need to treat every early Voice score as a scientific approval poll. They can treat it as a live signal from people paying attention to their actual decisions.

One angry pile-on may be noise. A repeated pattern across verified constituents, multiple Moments, issues, and time is harder to ignore. Voice is designed to separate those signals.

Your address tells Voice which politicians actually represent you. That makes your rating more meaningful: a rating from a verified constituent matters differently than a rating from a national spectator.

Voice uses your address to match you to districts like city, county, state, congressional district, and other relevant jurisdictions. Voice does not sell your street address, email, or personal identity to Wonk users, data partners, or any other outside customer.

Yes, where available and legally permitted, Voice can cross-reference a user's address with public voter-file records. This helps verify political geography and enrich aggregate analysis with fields that may include district, registration status, party registration where available, and turnout history.

Voter files do not show who someone voted for. They may show whether someone participated in certain elections. Voice uses this enrichment to improve weighting, detect duplicate or suspicious records, and understand whether ratings are coming from actual constituents, registered voters, frequent voters, new voters, or other relevant civic segments. Data made available through Wonk remains aggregated and anonymized.

No. Voice's data products are built around aggregated sentiment, not individual voter information. Wonk customers may license anonymized, aggregated ratings, trends, segments, and reports.

They do not receive individual emails, street addresses, names, or personally identifiable rating histories. The point is to make public sentiment louder, not expose individual voters.

They can try. Voice is built assuming politics is adversarial. Voice limits duplicate accounts, checks suspicious timing patterns, detects coordinated behavior, filters non-constituent activity, and separates local verified sentiment from national attention.

Voice does not pretend manipulation never happens. The job is to make manipulation harder, more visible, and less able to distort the public signal.

Early political reaction can still be useful if it is labeled honestly. Newsrooms show early vote counts before every precinct reports. Markets show prices before every possible participant trades.

Voice shows early sentiment as a developing signal, not a final verdict. Low-volume ratings are marked as early. Stronger signals are marked when more verified participation accumulates.

Signal strength is Voice's plain-English confidence label. It reflects rating volume, number of unique raters, constituent verification, geographic coverage, recency, and abuse checks.

A low signal does not mean a score is fake. It means the score is early. A stronger signal means more relevant people have weighed in and the pattern is harder to dismiss.

Public app ratings may be raw and unweighted unless labeled otherwise. Voice applies weighting when there is enough rating volume in a defined population to support a responsible model.

Weighting accounts for known differences between Voice participants and the population being described, using variables like geography, demographics, voter-file enrichment, registration status, turnout history, and other relevant benchmarks. When weighting is used, Voice says so clearly. When a score is unweighted, Voice says that too.

Weighting helps, but it is not magic. It can make estimates more useful and more comparable to a target population, but it cannot erase every bias. That is true for Voice, and it is true for traditional polling.

Voice publishes weighted estimates when the data supports it, discloses the method, and explains the limits.

Not automatically. People who rate politicians on Voice are likely more politically engaged than people who do not. That does not make the signal useless. It means the signal must be interpreted correctly.

Raw Voice ratings measure engaged public reaction. Weighted Voice estimates, when available, can better approximate a defined population.

Voice Moments are not designed to be blind ballot tests. They are designed to measure reaction to specific political information. That means sourcing, wording, and framing matter.

Voice cites sources, presents relevant context, avoids loaded language, separates facts from interpretation, and corrects errors quickly. A Moment rating is read as reaction to that Moment, not as a generic approval poll.

Journalists cite Voice data according to how the score is labeled. If a score is raw and unweighted, describe it as Voice user ratings or verified constituent ratings, not as a representative approval poll.

If Voice publishes a weighted estimate, the methodology is linked with the citation. Voice's goal is to make political sentiment easier to see without overstating what the data proves.

Voice discloses the score type, rating count, unique rater count, geography, recency window, constituent-verification status, whether the score is raw or weighted, and known limitations.

For formal reports and Wonk datasets, Voice provides fuller methodology: population definition, sample composition, voter-file enrichment fields, benchmark variables, weighting approach, quality controls, exclusions, and limitations. The standard is simple: if a number is public or licensed, readers know what kind of number it is.

Voice combines several safeguards rather than relying on one magic fix: voter-file enrichment to verify geography and improve weighting with public civic records where available; signal-strength labels to distinguish early reactions from stronger patterns; raw vs. weighted labels to prevent users from mistaking participation data for representative estimates; minimum-volume thresholds to avoid over-interpreting tiny samples; and constituent filtering to separate local verified ratings from national attention.

Voice also uses data-quality checks to detect bots, duplicates, suspicious timing, coordinated activity, and low-quality accounts; benchmark calibration to compare Voice patterns against known voter-file, census, election, or high-quality survey benchmarks where possible; occasional validation studies with controlled surveys or research partners; and transparent corrections when editorial or data errors are found.

Because politics needs more than slow, expensive snapshots. Polls ask a sample what they think. Voice lets constituents continuously react to what politicians do.

The honest version of Voice is not: "trust this number blindly." The honest version is: "here is the public signal, here is how strong it is, here is who it comes from, and here is what it can and cannot prove."