Methodology
Three Classes of Evidence
Everything published here belongs to exactly one of three classes, and we label which. (1) Observed public record: FEC filings, official House and Senate roll-call records, and Congress.gov sponsorship and cosponsorship records. (2) Documented third-party position: a position we can cite to a named, dated third-party source. We publish none of these unless the citation is present. (3) Automated classification: topical and industry categorization used to organize and navigate the data. Class 3 is never treated as a measured political position, and it is never mixed into a class 1 figure.
Recorded Votes
We publish only roll-call votes recorded in the official House and Senate record. Every row we display carries a chamber, congress, session, and roll-call number, and a position of Yes, No, Present, or Not Voting. Rows lacking those identifiers are rejected at both the ingestion and the render boundary.
Sponsorship Is Not a Vote
Sponsoring or cosponsoring a bill is a public statement of support recorded by Congress.gov. It is not a floor vote. Sponsorship and cosponsorship appear in their own clearly separated section and are never counted in any roll-call tally.
Why We Do Not Publish an Alignment Score
The public record documents how a member voted. It does not document how a donor or a district wanted them to vote. Any "donor-aligned" or "constituent-aligned" percentage would therefore be inferred rather than sourced, so we do not display one. People's Lobby does not use district polling, constituent polling, demographic proxies, or donor-interest heuristics to calculate alignment. No such figure exists anywhere in this product.
Roll-Call Coverage
House roll-call votes for the current Congress are ingested from the official clerk record via Congress.gov. Senate roll-call ingestion depends on the official Senate XML source, which is not reachable in every environment; where Senate votes have not been ingested, the member's receipt says the record is not yet processed rather than showing an empty or zero tally. Coverage expands as more of the record is processed.
Missing Data Is Not Zero
When we have not observed a record, we display an availability state — "not yet processed", "no matching FEC record", or "not enough documented data" — never $0, "none", or a blank chart. A zero is published only when a filing documents a zero.
Campaign Finance
Figures come from FEC filings for the current federal election cycle, which the app resolves dynamically rather than hard-coding. Contributions are reported in the categories the filing itself uses: PAC contributions, individual contributions, and small-dollar contributions under $200. We publish no qualitative funding labels such as "corporate-heavy" or "grassroots-driven", because those are characterizations rather than figures.
PACs vs. Employer-Linked Contributions
These are two different things and we never merge them. A PAC's total receipts and total disbursements are committee-wide FEC figures; they are not money given to any particular candidate, and we never label them "contributed" or "donated" to a member. Employer-linked figures aggregate contributions from individuals who reported that employer. They are not donations by the company, and the absence of such records never implies that a member is grassroots-funded.
ZIP and District Lookup
A ZIP code can span several congressional districts. If the lookup source cannot resolve an exact district, we return no House member at all and ask for a name or state instead of guessing.
Bill Translations
Legislative text is summarized into plain English using AI language models. These summaries describe effects only — they do not claim intent, assign blame, or use political framing. Summaries are class 3 material: an aid to reading the record, never a substitute for it.
Timeline
The timeline presents observed public events in chronological order: filings, committee assignments, and recorded votes. No causal connections are drawn.
AI Disclosure
AI is used only to summarize and translate public documents into plain English. AI does not score, accuse, rank, or predict behavior, and no AI output is used to compute any figure presented as public record.