Data Coverage

The data behind every Civly answer

Behind every report, donor list, and compliance check sits one unified data layer — the licensed national voter file, the full federal and state campaign-finance record, court and corporate filings, sanctions and adverse-action watchlists, and live social and media coverage. All primary-source, verified, and normalized so you can search them together.

2.6Brecords under one roof
128distinct data sources
50states in the voter file
1,500+data points per voter
200M+FEC contribution records
51.7Mstate legislative votes
10.7MU.S. court opinions
40+screening watchlists
14social platforms
7,272municipal meetings transcribed
4TV news channels captured 24/7
6.1Mfoundation grants searchable
358Division I athletics programs
What's in what service

Six services, one data layer

Civly is an AI intelligence platform for campaigns, committees, and partners. Every service draws on the same unified data — here's what each one powers.

The National Voter File

Inside every L2 voter record

Civly licenses L2, the national voter file — registered voters in all 50 states, with more than 1,500 data points on every record. It's the backbone of Civly's matching, contactability, and audience targeting. One national schema across all 50 states, so a list built in one state works the same way in the next.

What's on the recordWhat it covers
Registration & statusActive registration, party, and registration date for voters in all 50 states.
Vote historyWhether each voter turned out in past primaries and general elections, cycle by cycle.
Verified phonesCell and landline numbers appended to the record for direct contact.
AddressesCurrent mailing and residential addresses, standardized and mappable.
DemographicsAge, gender, and modeled race and ethnicity.
District assignmentsFederal, state, county, and local district for every voter.
Modeled partisanshipPolitical and turnout scores that rank likely support and likelihood to vote.
Wealth & consumer signalsEstimated income, home value, occupation, and education.
Lifestyle & interestsConsumer and lifestyle attributes for message and audience targeting.
Neighborhood & censusCensus and geographic detail around each household.
The Source Catalog

Every source, by category

128 data sources across finance, voting, court, wealth, influence, risk, media, nonprofit funding, and college athletics — grouped below into the families you would search them by, and normalized so you can search them together.

01

Money & Elections

Campaign finance
SourceWhat it is
Federal (FEC)Every federal contribution, committee, candidate, and expenditure. More than 200 million contribution records.
State & localCampaign-finance filings from nearly all 50 states, plus local jurisdictions — pulled directly from official government portals and refreshed automatically as new filings post. Where no statewide system covers a local race, county portals are ingested directly. Spending is covered alongside giving — Kansas expenditure filings are loaded, including the scanned paper filings, read column by column so payee names come through usable rather than mangled. See what union political money buys →
Behested paymentsPayments made at an official's request, on the public disclosure record. Statewide, plus the separate city filings for Los Angeles and Oakland.
Legislature & votes
SourceWhat it is
State legislation & votesBills, legislators, and the recorded floor votes of all 50 state legislatures51.7 million individual vote records across 1.19 million roll calls, covering 1.49 million bills, plus 184,000 indexed committee hearings.
Committee testimonyWho testified on a bill and on which side, from the hearing record. Live in Ohio, with more states to follow.
Floor amendments & motionsAmendments and tabling motions pulled from the chamber journals — the procedural votes that decide what a bill becomes and never appear in the official roll-call summary. Recorded votes on resolutions filed under a bill number, which the summary sources drop, are read from the journals as well. Live in Ohio, with more states to follow.
Federal bills & votesEvery federal bill since 2003 and roll-call votes, with a legislative-effectiveness score for each member.
Congressional RecordFloor statements by members of Congress, 1994 to the present — the largest body of attributable, on-the-record material a member leaves behind. Statements are bound to the member by official identifier, so a common surname never pulls in the wrong legislator, and each extract is labeled as entered into the Record rather than spoken aloud. The Record is edited, and members insert remarks they never delivered.
The electorate
SourceWhat it is
PollingAggregated public polls plus independent race ratings.
Prediction marketsLive implied win probabilities from Polymarket and Kalshi.
Election results & historyOfficial state canvass data for 2022 and 2024, covering every office on the ballot rather than Congress alone — 39,739 races and 131,005 candidacies across all 50 states and DC, drawn out of 30.7 million precinct-level rows. Candidate names are indexed, so what a person has won or lost is a lookup rather than a search. Multi-seat races are reported as multi-seat races, and write-ins and bookkeeping lines are not counted as candidacies. See which courthouse elections had no opponent →
Election schedulesThe forward-looking counterpart to certified results: which offices are up for election in a given state and year, in all fifty states. Five states publish an offices-by-year chart and were transcribed directly; for the rest the schedule is built only from checkable facts — federal offices, governor, and legislature, with legislative cycles derived from term length and whether a chamber turns over all at once or half at a time. County offices are read from the state constitution in Texas, from statute for Michigan's county commissioners, and for Colorado's county clerks. Limits are stated per state and per office rather than returned as a confident blank.
Ballot measuresCertified ballot-measure results computed per legislative district from the state's own canvass, with the passage threshold and precinct membership handled in code rather than worked out by hand. Live in Ohio, with more states to follow.
Political adsGoogle political-ad transparency: advertisers, spend, and creatives.
Local governmentCouncil and commission meeting records: hundreds of thousands of agenda matters and minutes.
Municipal meeting videoTranscribed, speaker-labeled video of 7,272 city and county meetings across 22 states, 2008–2026 — roughly 1.3 million searchable passages, each linked to the exact moment in the source video.
02

Capacity: wealth, assets & giving

Giving & wealth
SourceWhat it is
Charitable giftsMajor philanthropic gifts, university and museum honor rolls, and the Forbes 400.
Securities holdingsSEC filings: institutional 13F holdings, insider trades, private placements, and adviser records.
Congressional financesMember stock trades and net-worth estimates, plus the annual financial disclosures filed by both chambers, refreshed on a schedule. The 2025 Senate filings are in: 104 reports, 97 senators, 7,304 assets. Amended filings are kept alongside the original, so the figure you see is the amended one.
Foundation grants6.1 million grants read out of IRS filings and made searchable — 4,169,498 paid out by private foundations (990-PF) and 1,979,218 reported by the nonprofits that received them (990 Schedule I). Searchable three ways: what one funder has paid out, what one organization has taken in, and an open search by purpose, recipient state, amount band and year. Private-foundation contributors carry donor geography, enough to find the overlap between a subject's charitable giving and their political giving. See Grant Matching →
Who can apply, and by whenRead off the filings themselves: whether a funder accepts unsolicited requests at all — 69% of private foundations state that they only fund organizations they have already chosen — who to address a request to, the deadline, and the geography and field a funder restricts awards to. Where a foundation publishes application terms on its own website, those are read too, after the site is proved to belong to that foundation; that scan covers all 6,014 community foundations.
Open federal opportunitiesA live feed of federal grant notices, run alongside the filings because a tax return describes a cycle that has already closed: 1,690 notices, 537 of them open to a 501(c)(3), 97% carrying a hard closing date. Sortable by deadline, so what is about to close surfaces first.
Calls, conferences & awardsConferences, awards, pitch competitions and requests for proposals worth applying to, with the closing date read off the call page itself rather than inferred. Built from the noticeboards where solicitations are actually published, not only the individual calls a search surfaces, and refreshed weekly. Something passed on stops being suggested; something left unfinished returns once a new deadline is posted.
Nonprofit leadershipNonprofit officers, directors, and key-employee compensation from IRS 990 filings. Tens of millions of records. Three further schedules are read from the same filings: related organizations, transactions with interested persons, and political and lobbying activity.
Physician paymentsPharmaceutical and device-maker payments to physicians, from CMS Open Payments.
Public payrollsNamed public-employee salaries.
Federal disbursementsSBA PPP loans and USDA farm-subsidy payments.
Assets & ownership
SourceWhat it is
Corporate familiesGlobal legal-entity registry with parent and subsidiary links, plus state corporate registries — including New York, Colorado, Texas, and Florida — for business registrations, officer roles, and sole-proprietor status. Millions of entities.
UK companies & ownersThe UK company register and its beneficial-owner filings — the named people with significant control behind a British entity, not just the entity itself.
Property & county recordsReal-estate ownership plus county Registry of Deeds filings — deeds, liens, judgments, and foreclosures — matched to the subject.
Patents & trademarksThe federal intellectual-property record, 77 million records: granted patents and registered trademarks with the full ownership chain behind each, patent applications including the ones that never issued, roughly 12 million ownership transfers since 1970 naming both sides of a deal, and trademark disputes — who fought a subject over a brand, and whose brands the subject went after.
LiensSecured-lending (UCC) filings.
Aircraft & vesselsFAA aircraft and airmen, Coast Guard documented vessels, and live private-jet tracking.
Motor carriersFMCSA carrier registrations.
03

Legal, court & regulatory

SourceWhat it is
Court opinionsMore than 10.7 million U.S. court opinions, full text and searchable.
Dockets & judgesCase dockets, judge profiles, and judicial financial disclosures.
Judicial vacancies & appointmentsVacancies and appointments to the bench across eight states, with the election each appointed judge eventually has to face worked out from that state's own rules. Each refresh reports what changed since the last one rather than restating the whole board.
State civil & family docketsState-level civil and family court dockets, on a free-access track.
State human-rights commissionsDiscrimination and civil-rights complaints resolved at the state level — a record that sits outside the court system entirely.
Litigation lookupLitigation history tied to a person or entity.
EnvironmentalEPA facility enforcement and violations.
Workplace safetyOSHA inspections, violations, and accidents — down to the per-person injury detail behind an incident, not just the case summary.
Consumer complaintsThe CFPB consumer-complaint database. More than 14 million records.
Federal rulemakingFederal Register rules, notices, and presidential documents.
04

Influence: lobbying & contracts

SourceWhat it is
Federal lobbyingLobbying Disclosure Act filings, clients, and lobbyists. Nearly 2 million filings.
State lobbyingLobbyist registrations and expenditures across seven states, plus bill-level registrations — which employers lobbied a specific bill, cross-referenced against the subject's donor list. Bill-level coverage is live in Ohio.
Government contractsFederal contracts and grants, plus state procurement in California, New York, and Texas.
Federal registrantsThe SAM.gov registry of federal contractors and grantees.
05

Risk: screening & due diligence

SourceWhat it is
SanctionsThe U.S. Consolidated Screening List of twelve government lists, plus federal exclusions, screened in real time.
Adverse actionsHealth-care exclusions, FBI Most Wanted, FDA debarments, UN sanctions, and Federal Reserve and FEC enforcement.
ProsecutionsA searchable archive of Department of Justice press releases. More than 268,000 documents.
Document corpusScreening against the 383,000-document federal document release, with citations.
Extremist organizationsA reference list of 229 organizations, used to screen a subject's associations and endorsements.
Background & breachBackground-check signals and data-breach exposure checks.
06

Voice & Reach: social, media & narrative

SourceWhat it is
Social platformsCoverage across fourteen platforms: X, LinkedIn, YouTube, Instagram, TikTok, Facebook, Reddit, Bluesky, Substack, Truth Social, Snapchat, Rumble, Twitch, and Mastodon — with subject discovery reaching across the wider fediverse, not just the largest servers. Snapchat's pictures and video are read into words; TikTok reposts are collected and analyzed alongside original posts, which is often where someone shows what they endorse without writing a word of it; and Truth Social is read through the platform's own public API rather than a reseller.
Professional & company signalsHiring, expansion, funding, and leadership announcements from LinkedIn — the growth signals that say an organization is about to need something, pulled from the same feed as the mentions.
Podcast & radioAppearance monitoring across podcast and radio — including shows you can't name in advance — with each appearance verified against the source audio. After the fact, every clip cut from an appearance is traced back to whoever posted it, counted alongside the press pickups it drew.
TV newsFour national news channels captured around the clock and transcribed into searchable segments — every recorded hour watchable and clippable, with standing watches that surface a subject across the feed. Civly records and transcribes the broadcast itself, so coverage keeps running rather than depending on someone else's archive.
Video & audioDiarized transcription that labels who said what, so quotes are attributed to the right speaker. Where a platform ships video with no captions, the audio is transcribed rather than skipped.
Influencers & journalistsDiscovery and search for creators and reporters by topic and reach.
News & pressGlobal news and press coverage, current and archived.
Business reviewsPublic review and reputation signals.
AI narrativeWhat ChatGPT, Gemini, Claude, Perplexity, Grok, Google AI Overviews, and Google AI Mode say about a subject, graded across all seven.
07

Contact & identity enrichment

Turn a name into a verified way to reach the person.

CapabilityWhat it does
Verified mobileCell numbers from the L2 voter file — A+ identity matches carrying L2's strongest phone-reliability score, or the field ships empty.
Phone identity matchConfirms a phone belongs to the named person before you dial.
Email verificationDeliverability-checked business and personal email.
Address standardizationValidated, standardized mailing addresses.
Consumer contact appendA name and a verified street address turned into reachable contact information, through EnformionGO: phone numbers carrying the line type and whether the number is still connected, email addresses marked validated and split personal from business, the address history behind the current one, and a confidence score on the person the record belongs to. Built for lists that already carry a county-verified address, such as property records. Anyone who has asked EnformionGO to remove them is suppressed through the same list that carries the voter file's Do-Not-Call flag, so the two cannot drift apart.
LinkedIn profile matchThe confirmed LinkedIn profile behind a name — current title, employer, seniority, and tenure. The primary source behind lead scoring, B2B targeting, and professional capacity signals.
Professional contactsBusiness-contact enrichment for organizational outreach.
08

College athletics: money & athletes

SourceWhat it is
Athletic department financesRevenues and expenses per team from the federal Equity in Athletics survey every college taking federal student aid files annually, 2019–2025. Because it is a filing obligation rather than a records request, it reaches private universities that databases built on NCAA filings and state public-records law cannot.
Booster foundations & NIL collectivesThe nonprofits that fund a program, linked to the program they fund — a connection no public record makes. 63 organizations named across 54 of 358 Division I programs, $1.79 billion raised across their most recent IRS filings. Read the analysis →
Rosters & athlete productionSchool rosters and per-athlete season statistics, stored exactly as published so any figure traces back to its source. The inputs behind a NIL valuation and the school comparisons around it.
Recruiting & team strengthRecruiting classes and team-strength ratings for college football, and tempo-adjusted ratings for college basketball, loaded from published bulk files rather than a paid API. The 2026 class carries 4,453 recruits with 68 five-stars and a numeric grade on 79% of them; 2025 carries 4,171. The transfer portal is not in these files and is deliberately not built, because a partial portal reads as a complete one.
European professional rostersRosters from the European professional leagues — the half of a draft class that never appears in a college database.
Cross-source athlete identifiers10,902 identifiers linking an athlete to the outside databases — 6,627 with a Sports-Reference page, 3,493 with a RealGM page — each carrying the confidence of the match and the reason behind it. The resolver never accepts a name alone: it needs a second fact that independently agrees, such as birth date, birth city, or sport and school both matching. Only hard-evidence matches publish unqualified; the rest travel with their grade attached.
Program identityEvery Division I program's mascot and naming variants, so a collective named after neither its school nor its mascot still resolves to the program it funds — the join that surfaced booster foundations no existing list carried.
High-school eligibility & NIL rulesAll 51 state high-school associations are tracked. Where an association publishes its NIL rule on a web page, that page is checked daily — 16 states so far. Where the rule lives only inside a handbook PDF, the rule is extracted from the PDF; any state we cannot resolve is recorded as unresolved with the reason rather than left blank.
Athlete agent registersState registers of licensed athlete agents. Live in Texas, with more states to follow.
Fan forumsSchool message boards, resolved from the athlete’s own school rather than added one school at a time — so coverage travels with the subject. Where no board is known for a school, the report says so rather than reading as a quiet fanbase.
How the data gets clean

Every record earns its place

Public data arrives messy — a city field with a ZIP code sitting in it, a dollar amount that is really a date, the same donor spelled four ways in four states. Seven steps stand between the raw filing and the answer you get. Walk one contribution record through them.

Step 1

Capture it, and keep the original

Every source is pulled from the government portal or the record of origin — never a resold aggregator — and stored exactly as it was filed. All the cleaning that follows happens in a layer above that table, never on top of it.

  • The raw filing is never overwritten. Corrections live separately, so they can be reviewed, reversed, and re-run.
  • Rows we exclude from a total stay fully searchable in the source table. Nothing disappears.
  • Every number in a report traces back to the untouched row it came from.
Step 2

Clean every field

Dates, dollar amounts, names, and free text are parsed and checked on the way in. A value that cannot be real is emptied rather than guessed at — an empty field is honest, a wrong one is not.

  • An eight-digit number in an amount column is a date, not a $20,260,114 contribution. It gets rejected.
  • Dates outside a plausible range are dropped instead of landing a donation in 1902 or 2087.
  • Export artifacts — stray carets, null bytes, doubled spaces, trailing commas — are stripped out of names and text.
  • Titles typed into the first-name box (MR, DR, JR, III) are emptied, so no one becomes a person named “Mr.”
  • Placeholder employers (N/A, NONE, SELF) are emptied so they cannot climb a list of top employers.
Step 3

Fix the places

City and state are the dirtiest fields in public filings and the two people filter on most, so they get a pass of their own. Filers type an address into a city box; a state box collects everything from n.c. to nothing at all.

  • State codes are validated against the real list of states, territories, and military codes. Anything else is emptied, not kept.
  • ZIP codes, state abbreviations, country suffixes, and repeated names get stripped out of the city field: burnsvile nc 28714, becomes Burnsvile.
  • Redacted and placeholder cities — REDACTED BY SOS, N/A, SAME, VARIOUS — are emptied.
  • Then fuzzy matching finds the misspellings of a city that already exists and maps them onto the canonical name. Each correction is recorded in a table, so it is reviewable and applies to every future load automatically.
  • With deliberate exceptions, because a blunt rule does damage: Rocky Mt keeps its Mt.
  • Where a source ships the whole address as one blob and no city or state at all — the private-foundation giving file does exactly that — we parse them back out. 99% of that file now carries a valid state code it never shipped, which is what makes it filterable by geography at all.
Step 4

Group the free text

Occupation and employer are typed by hand by millions of filers, so the same job arrives spelled a hundred ways. Every distinct value is sorted once into a standard set of groups, so atty at law, Attorney, and Lawyer all filter as Legal.

  • Keyword rules handle the obvious cases, a trained model handles the rest, and a language model handles the long tail.
  • Each distinct value is classified once and cached, so the same text always gets the same answer everywhere it appears.
  • The filer’s original wording is kept next to the group. You can search either one.
Step 5

Link the records

One person turns up in dozens of filings across several states, spelled inconsistently every time. Resolving those into a single profile is the hardest step here, and the one that makes everything else worth doing.

  • Comparing every profile against every other would be roughly 500 billion comparisons, so candidates are narrowed first — by sound-alike surname, state, city, and initials — before anything is scored.
  • A trained statistical model then scores each pair and returns a calibrated probability that the two records are the same person, instead of a hand-tuned point score.
  • The bar is set high on purpose. For compliance work, a match we miss is far safer than two people wrongly merged into one.
  • The same machinery links across sources — the philanthropist on a foundation’s filing to the political donor, the licensed professional to the donor record, the booster nonprofit to the athletic program it funds. Those joins are the connections no single public record makes.
Step 6

Make it fast

Clean data nobody can query is not much use. The last mile is a serving layer built so a question that touches hundreds of millions of rows still answers while you wait.

  • 59 separate state and local contribution tables stack into one national table of more than 190 million records, then get snapshotted so a name lookup does not have to scan all of them.
  • Pre-computed summary tables sit behind every filter and autocomplete. On one 13.7-million-row table, listing the distinct cities went from 5.2 seconds to 5 milliseconds.
  • Indexes are built for the way people actually search: partial and misspelled text, most recent first.
  • Nightly, roughly 290 tables are written out as columnar Parquet files and queried with DuckDB. That is what lets a question spanning several sources at once return in seconds — and those same files are what a bulk export ships to your bucket.
Step 7

Guard it

All of the above is enforced by automated tests rather than by habit, because the failures that matter are the quiet ones — a bad row that makes a total look plausible but wrong.

  • A single contribution of $100 million or more is excluded from every donor total. Illinois once filed a $400,164,048 individual contribution — one typo that would otherwise have made that filer the largest donor in the entire database, ahead of every real megadonor in America.
  • That ceiling sits deliberately above the largest real contribution ever recorded, and a test fails if anyone lowers it into the range where genuine self-funders live.
  • A coverage test fails the build if a newly loaded state is not wired into the national view, so new data cannot go quietly missing.
  • And before research ships, a person reads it. AI accelerates the search; analysts confirm the facts.
How you get it

Any source, in the format you work in

Take any source as a report, a data feed, or an API call.

Platform

In-app search

Search and analyze every source from one secure hub, with autocomplete and filters.

One login
Documents

Decision-ready reports

Word and PDF deliverables, sourced and ready for debates, calls, and filings.

Word & PDF
Data

Structured feeds & API

JSON and the Civly Intelligence API for programmatic access to any product.

JSON & API
Bulk

Full data export

The entire research database delivered as Parquet to your own secure bucket.

Parquet export

Bring Civly to your work

Tell us the race, the list, or the vetting question, and we'll show you the exact sources that answer it. Request a demo of the platform, or an API key to build on the data directly.