Public money · US · USFED
US federal market intelligence
Every organisation that has received a US federal award since 2006, resolved to one identity, with its agency relationships and its month-by-month trajectory attached — scored for how much room there is to sell into it.
- One row per resolved organisation holding at least one US federal award.
- 733,144
- Award transactions from October 2006 to July 2026.
- Monthly
What this answers
Who the US federal government pays, how that has moved, and where a supplier or an investor should be paying attention. The transactions themselves are public; what is not published anywhere is a resolved organisation, and without one the same recipient appears under several spellings and several identifiers and no trajectory can be drawn through it.
The resolution is the product. The transactions are public.
The four things the score weighs
opportunity_score is a composite over four components, with the weights fixed
and published rather than tuned per edition:
- Award value percentile — 0.30
- Agency diversification — 0.25
- Recency, within 90 days — 0.25
- Twelve-month growth — 0.20
opportunity_drivers restates each row’s score in words — “1 agency, 19 awards,
active in the last 90 days, top 1.0% by value” — because a sales team will not
act on a number whose derivation it cannot repeat, and because a driver string
is what makes a wrong ranking arguable.
Note what diversification does in that formula. An organisation with one very large agency relationship and an organisation with six moderate ones can reach similar totals and land far apart on this score, which is the intended behaviour: concentration is a risk to the incumbent and an opening for everyone else.
Movers, and why they are a separate file
The ranked accounts file answers “who is big”. The movers file answers “what changed”, at one row per organisation with a material 90-day shift in obligations. They are separate because they are read at different cadences — the first quarterly by a strategy team, the second monthly by whoever owns the pipeline — and because one combined file would be sorted by the wrong column for one of those two readers.
What travels with it
Six files: the ranked accounts, the movers, the organisation entity record, the organisation-by-agency and organisation-by-month relationship tables, and the identifier crosswalk. Award transactions at their own grain are part of the top delivery tier rather than of every edition. The complete column dictionary is below.
What it does not claim
Stated before the sample rather than after them. Everything below is a measured property of the file, not a legal hedge — if one of these is the reason the file will not work for you, that is worth knowing in the first five minutes rather than the fifth week.
- 733,144 is the count of organisations that have actually been awarded something. It is not the count of registered entities: 1,767 further organisations carry a federal registration identifier and have never received an award, and they are deliberately absent.
- The opportunity score is relative within an edition. It ranks organisations against each other in this cut of the data and is not comparable across editions after a schema change, which is stated on the score itself rather than in a footnote.
- Obligations are summed per award rather than per transaction row, and a negative total is a real outcome where de-obligations exceed obligations. It is not a data error and should not be filtered out without a reason.
- Agency counts are by slugged buyer name. Where a department reorganises and renames, the trajectory of a relationship can break across the rename, and nothing in this file stitches it back together.
- This edition's figures were measured by reproducing the tables the product reads from the real Silver with the build's own code, rather than from a completed production build. The declared population was corrected downward as a result, from 734,911 to 733,144.
Preview
Five rows from the 2026.09 build, cut on 25 August 2026 against the fixture lake — so the shape, the columns and the ordering are exactly what ships, and the values are not the production population. Identifying columns are withheld from this preview.
| rank | opportunity_score | opportunity_drivers | agency_count | last_award_date |
|---|---|---|---|---|
| 1 | 77.3 | 1 agency, 3 awards, active in the last 90 days, top 1.0% by value | 1 | 2023-08-16 |
| 2 | 77.2 | 1 agency, 19 awards, active in the last 90 days, top 1.0% by value | 1 | 2023-08-21 |
| 3 | 77.0 | 1 agency, 21 awards, active in the last 90 days, top 2.0% by value | 1 | 2023-08-21 |
| 4 | 77.0 | 1 agency, 6 awards, active in the last 90 days, top 2.0% by value | 1 | 2023-09-20 |
| 5 | 76.7 | 1 agency, 9 awards, active in the last 90 days, top 3.0% by value | 1 | 2023-09-20 |
The full sample pack carries every column, roughly a hundred rows per file, the methodology note and the licence. Ask for it against the ID.
Request the USFED sample pack →Column dictionary
Generated from the built release, not written by hand. A column that passes through unchanged inherits its definition from the publisher's own contract; a column we derive carries prose written against the query that produces it. Both directions reconcile — every column produced has a definition, and every definition names a column that is produced.
top_accounts
- org_id Our resolved organisation identifier. Stable across editions and the join key for every file in this pack.
- rank Position by opportunity_score within this edition. Not comparable across editions.
- legal_name Name as we resolved it, taken from the strongest register the organisation appears in.
- opportunity_score Composite 0-100. Weights: award value percentile 0.30, agency diversification 0.25, recency 0.25, 12-month growth 0.20. Relative within this edition.
- opportunity_drivers The reasons behind this row's score, in words.
- total_obligation_usd Sum of obligations over distinct awards, US dollars. Negative where de-obligations exceed obligations, which is a real outcome and not an error.
- agency_count Distinct awarding agencies, by slugged buyer name.
- last_award_date Most recent award from this agency.
- release_id The release this row was published in. Quote it in any support question and we can name the exact build it came from.
movers
- org_id Our resolved organisation identifier. Stable across editions and the join key for every file in this pack.
- legal_name Name as we resolved it, taken from the strongest register the organisation appears in.
- recent_obligation_usd Obligations in the most recent 90 days of the edition.
- prior_obligation_usd Obligations in the 90 days before that.
- activity_delta_90d Recent minus prior, US dollars.
- direction `entered`, `accelerating` or `decelerating`.
- first_award_date Earliest award held for this organisation.
- release_id The release this row was published in. Quote it in any support question and we can name the exact build it came from.
organisations
- org_id Our resolved organisation identifier. Stable across editions and the join key for every file in this pack.
- legal_name Name as we resolved it, taken from the strongest register the organisation appears in.
- postcode Postcode of the registered address, as the resolved spine holds it.
- uei The US federal supplier entity identifier.
- uei_count How many SAM registrations resolve to this organisation. Above one means the spine merged several registrations; a large number is worth checking before quoting the total.
- award_count Distinct awards from this agency to this organisation.
- total_obligation_usd Sum of obligations over distinct awards, US dollars. Negative where de-obligations exceed obligations, which is a real outcome and not an error.
- agency_count Distinct awarding agencies, by slugged buyer name.
- first_award_date Earliest award held for this organisation.
- last_award_date Most recent award from this agency.
- counterparty_hhi Herfindahl-Hirschman index of obligations across agencies, 0-10,000. 10,000 means a single agency.
- top_counterparty_share Share of total obligation from the single largest agency, 0-1.
- uncompetitive_awards Awards the publisher states were not competed. Null where the publisher is silent, never false by default.
- release_id The release this row was published in. Quote it in any support question and we can name the exact build it came from.
org_agency
- org_id Our resolved organisation identifier. Stable across editions and the join key for every file in this pack.
- agency_id Deterministic slug of the awarding agency name. Not a federal identifier: the published data carries no agency dimension.
- agency_name The awarding body as the publisher names it, normalised to one spelling per buyer.
- award_count Distinct awards from this agency to this organisation.
- obligation_usd Sum of obligations over distinct awards, US dollars.
- first_award_date Earliest award held for this organisation.
- last_award_date Most recent award from this agency.
- supplier_count How many organisations this agency awarded to, across the whole edition. A low count on a large agency is a concentration signal.
- release_id The release this row was published in. Quote it in any support question and we can name the exact build it came from.
org_month
- org_id Our resolved organisation identifier. Stable across editions and the join key for every file in this pack.
- month First day of the calendar month of the award action date.
- award_count Distinct awards from this agency to this organisation.
- obligation_usd Sum of obligations over distinct awards, US dollars.
- release_id The release this row was published in. Quote it in any support question and we can name the exact build it came from.
crosswalk
- org_id Our resolved organisation identifier. Stable across editions and the join key for every file in this pack.
- legal_name Name as we resolved it, taken from the strongest register the organisation appears in.
- name_normalised Upper-cased, punctuation stripped, whitespace collapsed. For fallback matching where no identifier is shared.
- uei The US federal supplier entity identifier.
- lei Legal Entity Identifier, where we hold one. Cross-border and investor joins.
- sec_cik US securities filer identifier, where we hold one.
- ein US employer identification number, where we hold one.
- companies_house_number UK company number, where the organisation is also on the UK register.
- country Jurisdiction of the sources that contributed, not the organisation's trading country.
- release_id The release this row was published in. Quote it in any support question and we can name the exact build it came from.
Licence and delivery
- CSV · Parquet
- Presigned download of a single archive, with the release ID stamped into every row.
- None. The population is organisations, and no individual recipient or officer column reaches this file.
- Every upstream source is open, permits redistribution and permits resale — which is why this file is in the catalogue at all. The clearance is checked column by column at build time, and a file that fails it is not cut.
- Named in the sample pack and in the delivered files, with the attribution each licence requires, rather than on this page. Which registers combine to make a product is the part of this work we do not publish.
Enquiries
Quote USFED at data@inference.institute. It is the same code the build and the release manifests use, so an enquiry carrying it lands against exactly this product.
Next step
Evaluate it before you commit to it.
The sample pack has the real schema and a watermarked extract, so the question of whether this file answers yours is settled with data rather than with a call.