Methods and corrections

The exact rules behind every grade, share and label on this site, and how to tell us when we got something wrong. Last updated October 28, 2026.

Outside money grades

Every group that spent $100,000 or more on independent expenditures (ads and outreach for or against a candidate, without coordinating with the campaign) in 2026 House and Senate races gets a score from 0 to 100. The formula is the same for every group, whichever party or candidate it helps. A grade measures how much voters can see about who pays for a group's ads. It is not a finding that anyone broke the law; most of what lowers a grade is legal.

Traceable share (60%)

Each dollar a group reported raising gets credit by how far it can be followed to a person or a named organization:

A group that reports ads to the FEC but is not a registered committee files no donor list, so it is judged by what kind of organization it is: a union or company gets full credit, a nonprofit none.

On-time share (25%)

Ads that cost $10,000 or more must be reported within 48 hours of going public, and $1,000 or more within 24 hours in the last 20 days before an election. We take every ad dollar large enough to need one of those reports and count the share that first reached the FEC within 2 days of the date the group said the ad went public. A group with no ads that needed a fast report gets full credit.

FEC letters (15%)

FEC analysts send a Request for Additional Information when a report looks incomplete, inaccurate or late. The letters measure is 1 minus (letters this cycle ÷ reports the group filed with itemized activity), never below 0. A group with as many letters as reports scores 0 on this part. If the letters file is unavailable when the site is rebuilt, the score uses the other two parts reweighted to 100.

Letters are listed newest first and counted against an hourly limit on FEC API calls; if that limit is reached mid-count, the site still publishes rather than wait, using whatever letters it read before stopping. An undercount can only make this part of a score look better than it should, never worse, since a missed letter is one fewer counted against a group. When this has happened for the current scores, it's noted where the letters measure appears on a group's page.

Edge cases

• Amended reports. The newest version of a report replaces the older one. An ad listed again on an amended report under a new transaction ID is counted once.
• Name matching is done by computer, so a company or nonprofit can occasionally be misclassified. That is the most common reason to dispute a grade, and the quickest to fix.
• Timing. A group that raised money late in a reporting period may show ad spending before the matching receipts appear; its grade can rise when the next report is filed.
• Refresh. Grades are recomputed from fresh FEC data at every twice-weekly rebuild. A grade can change without anyone asking.

Race and state transparency shares

For a race, the untraceable share is outside money spent in that race times each spender's untraceable share (1 minus its traceable share), summed over spenders, divided by all outside money in the race. States use the same calculation over all of their federal races. Groups below the $100,000 grading threshold are counted with their own traceable share even though they get no letter grade.

Who funds the ads

On race pages and in the outside-spending lists, each graded group shows where its money came from:

• If one committee (another super PAC, a party, a joint fund) supplied at least half of a group's money, we follow it, and repeat up to three steps: "Funded through $158M from SLF PAC (Senate Leadership Fund), whose largest donors are..."
• At the end of that trail, or for the group itself when no single committee dominates, we list up to three people who gave $1 million or more to that committee, with city and state. Smaller donors are never named here; they remain in the donor search.
• The amounts are what each person gave to that committee. Money is not earmarked in FEC reports, so we do not claim a specific donor's dollars paid for a specific ad; we show who funds the committee the money came from.
• The same rule applies to every group, whichever party or candidate it helps.

The header search also finds everyone who gave $1 million or more this cycle, and committees by the names people use for them (for example, Senate Leadership Fund for SLF PAC).

Which election an ad counts toward

An ad counts toward the election it ran before. We use the date the group says the ad was first shown, or, when it gives none, the date it paid for the ad. Ads that ran before January 1, 2025 belong to the 2024 election and are not counted, even when the group amended its report later. The FEC files an amended report under the year it was filed, so a 2024 ad on a report amended in 2025 arrives labeled 2026. Ads with neither date are kept.

The rule is the same for every group and every page: race totals, state totals, a group's own totals, its map and its grade. Every rebuild prints how much the rule removed, and stops if that is more than $25 million or 2% of all outside spending, because that would mean the dates were misread rather than that groups amended old reports.

Reading a group's or donor's map

• One circle per race. Everything a group put into one Senate race or one House district is one circle: its money for every candidate there and against every candidate there. Senate circles sit at a point inside the state (the point farthest from its edges). House circles sit on a point inside their district (see district points). The tooltip, the list and the breakdown you get by clicking a circle split each one by candidate, for and against.
• Size is dollars. A circle's area is proportional to the money, on one scale for the whole map, so a circle twice the area is twice the money. Circles grow as you zoom in, all by the same factor, so their sizes still compare; the legend shows two reference amounts at the current zoom. A deeper shade also means more money, on the same scale. Larger circles are drawn first, so a smaller one on top of a larger one stays visible and can be clicked. Amounts under $1,000 are listed but not drawn.
• Color is whom the money helps. Ads or gifts for a Democrat, and ads against a Republican, help Democrats; the reverse helps Republicans. A race where all the money helps Democrats is blue, all Republicans red, and a mix is a blend through purple, in proportion. Ads against someone in a primary help a rival from their own party, but only when they had one (another candidate of their party registered for 2026 in that race); otherwise the ad counts against their party like any other.
• Gold is independents. Money that helps an independent or third-party candidate (for them, or against their opponent when no candidate of the other major party is on the ballot) gives the circle a gold ring when it is 10% or more of the race's money; the ring is wider for a bigger share. When an independent gets most of the money, the whole circle is gold. Smaller shares get no ring, but the tooltip and the breakdown always give the exact figure. The fill otherwise shows the Democratic and Republican blend.
• The filters apply to the map and the list together. By default, candidates not on the November 3 ballot are left out; a checkbox brings back their primary spending. The line under the filters always says how many candidates and dollars are shown.
• Why the map can be larger than "Spent". A group's Raised and Spent figures come from its latest full report. The map also counts ads reported since then on 24- and 48-hour notices, and the page says how much. The "Since its last full report" view shows only those.

District points

The map does not draw district outlines; it places each House race's circle on one point inside the district in use on November 3, 2026: the point farthest from the district's edges on land (a pole of inaccessibility), so even an oddly shaped district gets a point inside it. The points are longitude and latitude read straight from each state's official boundary file. Circles are never moved off their points: in a dense area such as New York, Los Angeles or Chicago, zoom in and they separate.

• New 2026 lines, from each state's official files: California (California State Assembly, AB 604 (Proposition 50), 2025-11-04), North Carolina (North Carolina State Board of Elections, S.L. 2025-95, 2025-10-22), Texas (Texas Legislative Council, PLANC2333, 2025-08-29), Utah (Utah Geospatial Resource Center, US Congress Districts 2026 to 2032 (court-adopted Map 1), 2025-11-10).
• Lines unchanged since 2024, from the U.S. Census Bureau's 119th Congress districts: AK, AR, AZ, CO, CT, DE, GA, HI, IA, ID, IL, IN, KS, KY, MA, MD, ME, MI, MN, MS, MT, ND, NE, NH, NJ, NM, NV, NY, OK, OR, PA, RI, SC, SD, VA, VT, WA, WI, WV, WY.
• Placed approximately, near the state's center, because the 2026 lines could not be confirmed from an official boundary file: Alabama (new map passed in the May 2026 special session; its status for 2026 is reported both ways, so no lines are assumed); Florida (new map signed May 4, 2026 (plan EOGPCRP2026); only PDF maps were found from the Legislature); Louisiana (new map enacted in the 2026 regular session (Act 2, SB 121), under challenge; no district boundary file was found); Missouri (the 2025 map was blocked and revived in court through September 2026; which map is in use could not be confirmed); Ohio (new map adopted by the Ohio Redistricting Commission Oct 31, 2025; the commission publishes a block assignment file, no boundary file was found); Tennessee (new map enacted May 7, 2026; only a PDF map was found from the Comptroller). These states' House circles are placed near the state's center, spread out so they don't stack, and each one's breakdown says its spot is approximate; they move to their districts once an official file is found.

As of 2026-10-28. The points are rebuilt by hand with build_districts.py when a state's lines change. The map is drawn with MapLibre GL JS, on U.S. Census Bureau state and county boundaries (us-atlas), all served from this site.

Primary results and the November ballot

The FEC lists everyone who registered to run, including candidates who lost a primary, withdrew or never qualified. To keep race pages focused on the actual contest, we mark who is on the November 3 ballot, one race at a time:

• Nominees come from each state's 2026 election results as published on Wikipedia's state election pages, Ballotpedia, The Green Papers or state election offices, with one source recorded per race.
• A race is marked confirmed only when every nominee found is matched to an FEC registration in that race, or has been checked by hand as a nominee who never filed with the FEC (usually because they raised too little to have to). If a nominee could plausibly be one of several FEC filers of the same party, the race stays unconfirmed, and the race page shows every registered candidate with a note saying so.
• In a confirmed race, every other registered candidate is labeled "not on the November ballot" with the state's primary (and runoff) date. That label covers losing a primary or runoff, withdrawing, and never qualifying; it does not say which.
• Louisiana's House races hold their first round on November 3, with a December 12 runoff (the party primaries were canceled after the congressional map changed), so they have no nominees and are shown unconfirmed.
• A few FEC registrations name a district that does not exist (for example Georgia's 23rd). Where we found the race the person actually ran in, they are labeled by that result.
• Money a candidate raised before leaving the race still counts in state and national totals, because it was raised for this cycle.

State disclosure-law grades (pilot)

Thirteen tests worth 100 points grade how much a state's laws let voters see who pays for outside spending in state elections: outside spending reported (30 points), issue ads near elections reported (20), donors behind the spending disclosed and traced (30), and speed and public access (20). Each test gets full, half or no credit, and every point cites the statute or rule. Same A to F scale as above. Seven grades so far (six states and the federal rules). These are a pilot: an election lawyer has not reviewed them yet, and two New Jersey findings are marked not fully confirmed on the page.

Insights page

The Insights page runs 16 automatic tests over the site's own published data and lists patterns worth a closer look. A lead is not an accusation: each one shows the evidence, why the pattern can matter, and the ordinary explanations. No individual donor is ever named there. The tests rerun with every twice-weekly rebuild.

Within each test, leads are ranked by a severity score that grows with the size of the money and how far past the threshold it is. At most 80 leads are kept per test (60 for clusters, repeats and late money), strongest first. Self-funders, debt and out-of-state money are listed as interesting rather than stronger signals.

What stands out

Each candidate's page opens with three sentences on where that campaign differs most from its opponents. The same rule picks them for every candidate; no one chooses them by hand.

• Opponents. The other nominees in the same race on the Nov 3 ballot. When a race is not confirmed yet (Louisiana's House races), when a nominee has no opponent who has filed with the FEC, or for a candidate who is not on the ballot, the comparison is with the typical campaign for that chamber: the median of Senate or House campaigns on the ballot that raised $250,000 or more.
• Measures. Money raised; the shares of it from small donors ($200 or less), from donors in the home state (of named donors' money), from PACs and other committees, and from the candidate; outside money for and against; outside money for, divided by what the campaign raised; the share of the campaign's own itemized spending in each spending category; food and drink per $100 spent; and large donations reported on 48-hour notices since Oct 15.
• Size of a difference. For money and ratios, the gap is the absolute logarithm of one value over the other, after adding $10,000 to money and 0.05 to ratios so that very small numbers don't dominate. For percentages it is the difference in points, divided by 10. A measure counts the largest gap between this candidate and any opponent.
• What doesn't count. A gap under 10 points, under $100,000 in a Senate race or $25,000 in a House race, or a comparison where either value is missing. A measure counts only when this candidate has the highest or the lowest value of everyone compared, and not when this candidate's own value is zero.
• Small campaigns in spending categories. For the share of spending in each category, opponents who raised under $100,000 are left out of the comparison: in a campaign that small, one fixed cost such as a filing fee can be 40% of all its spending. They stay in every other comparison, and the sentence says when one is left out.
• Picking three. The three largest gaps, with at most one spending category and at most one outside-money measure.
• Wording. Each sentence gives this candidate's figure first, then each opponent's, without adjectives. Every sentence has a bar chart under it showing all the nominees.

Food and drink

A payment counts as food and drink when its amount is positive, its stated purpose mentions meals, food, catering, a caterer, dinner, lunch, breakfast, beverages, coffee, a restaurant, dining or refreshments, and the purpose does not also mention reimbursement, mileage, travel, a venue, a facility, event space or a hall, a site fee, a rental or rent, golf, an outing, T-shirts, parking, lodging, a hotel, accommodations, airfare, gas, transportation, a car service, supplies, tickets, consulting, commissions, staffing or personnel, media, advertising, printing, postage, shipping, phone, software, email, data, dues, membership, entertainment, gifts, mementos, "see below" or in-kind, since those payments bundle food with something else. Payments to a party committee are left out too: a party's dinner is dues or a sponsorship. Purchases listed under a campaign's credit card payments count on their own. Food billed by a payee whose spending is mostly consulting, fundraising, digital, mail or media work is shown on its own line ("also billed through vendors") rather than in the total. The figure per $100 divides the total by the campaign's itemized spending.

Corrections and disputes

If a figure is wrong because of what was filed (a misspelled name, a wrong amount, a wrong employer), the correction has to come from the filer: campaigns and committees amend their reports with the FEC, and this site picks up the amended report on its next refresh, usually within a few days.

If we made the mistake (a group misclassified, a payment in the wrong category, a wrong primary label, a law misread), tell us and we will fix it. What helps:

• The page link and what is wrong.
• For a grade: which source we misclassified (for example, "the $2 million from X is a company, not a nonprofit"), with a public record that shows it.
• For a primary label: a link to the official results.

We review every request, answer within five business days, and fix confirmed errors at the next refresh. Grades follow the formula above; we do not change a grade by request, only the facts that feed it. Significant corrections are listed below with the date.

Corrections log