Start a conversation Contact

Data · licensable

Data you can check before you buy it.

Company registers, property titles and public spending, resolved to one organisation and cut into files with a schema and a licence. Every column on every file is defined and every figure traces to a measurement with a date on it. 22 packaged, and 66 more specified and ready to commission.

5 available · 17 in build · 22 in the catalogue

UKREGCON UK registered-office concentration Every UK company set against the other companies registered at the same address, with the cohort's dissolution and overdue-filing rates attached — so a mass-registration address, a formation agent and an ordinary serviced office can be told apart. Risk & conduct UK 2,822,405 organisations Available UKASSET UK filed asset base and charge coverage Every active UK company with a readable filed asset figure, broken down by asset class and set against the charges registered over it — so the question "what is actually behind this company's balance sheet, and who already has a claim on it" has one row. Corporate UK 2,801,757 organisations Available 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. Public money US 733,144 organisations Available UKCREMAP UK commercial property landlords, ranked by distress Every UK company holding ten or more registered property titles, scored on what the public record says about its condition — balance-sheet insolvency, current ratio, overdue filings and insolvency notices — with the portfolio and the reasons behind the score attached. Property UK 13,269 organisations Available UKCRE UK corporate landlords carrying both a charge and a distress signal The screened half of the landlord population: companies holding ten or more registered property titles that carry at least one outstanding charge and at least one filed or published sign of difficulty. A shortlist rather than a market map. Property UK 4,686 organisations Available
UKUSFLEET Commercial vehicle operators, UK and US Goods and passenger operator licences joined to US carrier registrations, as one view of who runs commercial vehicles on both sides. Outstanding · Waiting on both licence registers reaching the lake. Effort rather than readiness — the cheapest of this wave to finish. Corporate UK · US Coming soon Register interest
UKDEADCO Dormant in name only UK companies whose register status and filing conduct disagree — dormant or active on paper, and behaving otherwise in the filings. Outstanding · Waiting on two upstream inputs reaching a readable table. Corporate UK Coming soon Register interest
UKFIN UK company financials Filed accounts facts for every UK company that files them, on one concept vocabulary, with the filing history behind each figure. Outstanding · Waiting on confirmation that the bulk accounts product is the same publisher product already under contract. Corporate UK Coming soon Register interest
UKGROWTH UK growth signals Sponsor-licence grants, share allotments, hiring velocity and movement in the filed accounts, combined into one growth signal per UK company. Outstanding · Waiting on document extraction, and on a measured error rate for it. The share-allotment half is the distinctive part and does not exist until then. Corporate UK Coming soon Register interest
USUKCHAIN US/UK cross-border corporate presence Organisations carrying a registration on both sides of the Atlantic, resolved to one identity with the evidence for the match. Outstanding · Builds green; waiting on the rebuild that publishes its input table. Corporate UK · US Coming soon Register interest
USUKCONC US/UK record concentration Where one organisation accounts for an outsized share of the records held against a register, an address or a buyer. Outstanding · Builds green; waiting on the rebuild that publishes its input table. Corporate UK · US Coming soon Register interest
UKPUBMONEY The UK public money graph Public bodies, the organisations they pay, and the awards and grants that connect them, on one resolved spine. Outstanding · Builds green; waiting on the rebuild that publishes its input table. Public money UK Coming soon Register interest
UKPSCONC UK public sector supplier concentration Buyer markets where a small number of suppliers hold most of the awarded value, measured per buyer rather than nationally. Outstanding · Builds green; waiting on the rebuild that publishes its input table. Public money UK Coming soon Register interest
USAWARDS US federal award concentration Recipients of US federal awards ranked by concentration of obligations across agencies and time. Outstanding · Builds green; waiting on the rebuild that publishes its input table. Public money US Coming soon Register interest
USCONINTEG US federal contractor integrity Federal award recipients joined to the exclusion and debarment registers, with the match evidence travelling alongside. Outstanding · Builds green; waiting on the rebuild that publishes its input table. Public money US Coming soon Register interest
USHEALTHPAY US healthcare payments Reported industry payments joined to the provider registry and to nonprofit hospital networks, per recipient and per payer. Outstanding · Two things, and the second is the harder one. The payments source is not held, so there is nothing yet to join. And it carries personal data about named individuals, which is a live regulatory exposure that counsel has not answered. Public money US Coming soon Register interest
UKCTRL Controller flight risk The screened half of the phoenix population: a single delivery naming controllers who meet the pattern, rather than the map it is drawn from. Outstanding · The same assessment the phoenix map is waiting on. It is the shortlist cut of that file, not a separate measurement. Risk & conduct UK Coming soon Register interest
UKUSXBORDER Cross-border structure signals UK companies whose declared controller sits in a US state with light disclosure, and whose filed accounts say dormant while the balance sheet does not. Outstanding · Waiting on the same subject-axis assessment as the other control-register products, and on the US half being resolved to the same spine. Risk & conduct UK · US Coming soon Register interest
UKESGRISK Environmental claim against environmental record The gap between what a UK company told the state it does — its declared activity, its licences — and what a regulator recorded it doing. Outstanding · Waiting on a documented legitimate-interest assessment: both regulatory inputs carry individuals as well as companies, so they are withheld on the subject axis rather than the licence. Risk & conduct UK Coming soon Register interest
UKFLIGHT Phoenix companies, while the liquidation is open Controllers of three or more failing UK companies who have already incorporated a replacement in the same trade — found while the insolvency is still running rather than after it closes. Outstanding · Waiting on a documented legitimate-interest assessment: the control register is individuals as well as companies, so the input is withheld on the subject axis rather than the licence. Risk & conduct UK Coming soon Register interest
UKEMPRISK UK employer conduct record Tribunal outcomes, health-and-safety enforcement, minimum-wage naming and pay gap reporting, resolved to one view per UK employer. Outstanding · Most inputs are held. It carries personal data — tribunal decisions name claimants — so an employer-level file has to aggregate to the respondent and publish no claimant. That reframing is not decided. Risk & conduct UK Coming soon Register interest
UKIPLEADS UK insolvency practitioner appointments Published insolvency notices resolved to the appointed practitioner and the firm behind them, with the appointing company matched on name. Outstanding · Builds green; waiting on the rebuild that publishes its input table. Risk & conduct UK Coming soon Register interest

Build to order

66 more are specified, and not yet cut.

Each of these is written down: the question it answers, what it is built from, and what the answer looks like. None of them is a file sitting on a shelf, and none of them is presented as one — but the inputs are held, the method is settled, and 33 of the 66 are a matter of days rather than months.

Commissioning one is the same conversation as licensing a packaged dataset, quoted against the work rather than the file. Quote the code.

Commission one →

Aggregation and benchmarks · 14

  • D1 UK Sector Financial Benchmarks “Is this company's gearing normal for a mid-sized construction firm in the North West?” P10/25/50/75/90 per concept × SIC section × size band × region × year, min cell 5 Days
  • D10 Insolvency Rate Panel “Is my sector's failure rate rising?” Notices per 1,000 live companies by type × sector × region × month Days
  • D11 Environmental Compliance Benchmark “Is this operator's enforcement record unusual for its permit class?” Actions and fine amounts per operator class; fine_amount and costs_amount are published Days
  • D13 Fleet and Operator Benchmarks “How does this haulier's fleet and safety record compare?” Fleet-size bands, authorised-versus-specified ratio, safety-rating distribution by state and licence type Days
  • D14 Payment Behaviour Benchmark “Who pays late, and by how much, in this sector?” Days-to-pay and %-paid-late distributions by sector × size × reporting period Days
  • D2 Romanian Sector Benchmarks “The same question, with a real income statement” As D1, plus margin and expense-ratio bands GB cannot support Days
  • D4 Public Procurement Market Sizing “How big is this market, who buys, and is it growing?” Spend and count by buyer × CPV/NAICS × year × country; 92.9 M US and 33.6 M RO awards carry a value Days
  • D5 Buyer Concentration Index “Is this buyer's supply base concentrating, and around whom?” HHI and top-N share per buyer × category × year across 304 US, 22,590 RO and 7,687 GB buyers Days
  • D7 Neighbourhood Business Panel “What kind of place is this, commercially?” Business density, premises mix, IMD decile, tenure and qualification rates per postcode sector and LSOA Days
  • D8 Mass-Registration League Table “Which addresses host implausible numbers of companies?” Measured: 32,289 addresses are already classified mass_registration, against 988,864 commercial and 27,331,734 dwellings Days
  • D12 US Wage Floor Benchmarks “What must we pay for this role in this metro to clear a certification?” Prevailing-wage percentiles by SOC × MSA × year, plus filed-wage distribution from 9.9 M LCA rows Weeks
  • D3 Cross-Border Benchmark Pack “How does our Romanian subsidiary compare with the UK parent's peer set?” One concept vocabulary across three countries; currency-normalised Weeks
  • D6 Competitive Tension Index “Which categories are competed and which are not?” 38,953,328 US awards record competition and 24,713,162 RO ones do; single-bid rate by buyer × category × year Weeks
  • D9 UK Company Demography Panel “What is the birth, death and survival rate of firms like ours?” Cohort survival curves by SIC × region × year; hazard rate by age band Weeks

Enrichment and entity resolution as a service · 8

  • D33 Canonical Financials Feed “Give me GB, RO and US accounts on one schema.” The concept mapping itself is the product — 22 GB and 18 RO concepts reconciled, with measure_kind, period_plausible and is_latest flags Engine built
  • D27 Identifier Crosswalk API “What is this company's LEI / CIK / UEI / CUI / IČO?” The open crosswalk sells as data (F4); the full crosswalk answers as a service, per row, with per-row provenance (declared / derived / matched) Days
  • D28 Company File Enrichment “Append everything you know to my 100,000-name customer list.” The 82 columns of company_profile_open, appended, with a coverage report per column Days
  • D29 Ultimate Owner Append “Who ultimately controls this company?” Measured: 723,212 companies with a corporate owner, 260,079 distinct ultimate owners, max chain depth 33, 3,101 hitting the depth cap, 1,232 in a cycle, 46,189 with multiple controllers — all four disclosed per row Days
  • D31 Supplier List Matching “Match my 20,000 suppliers to the award record.” Resolution plus award history per matched supplier, returned as an aggregate per supplier Days
  • D32 Classification Crosswalk Service “Map my portfolio between NACE revisions and national schemes.” Official correspondences: NACE Rev 2 ↔ Rev 2.1, TEÁOR'08 → TEÁOR'25, SIC 2007. Publisher riders require derived data to be marked as derived Days
  • D26 Entity Resolution API “Which company is this, really?” The published scoring model: identifier 1.00, exact name 0.55, token similarity 0.35×sim, postcode 0.30, town 0.25, threshold 0.80 — a name alone never reaches it; registry-primacy schemes veto Weeks
  • D30 Address Normalisation Service “Are these two addresses the same place?” Normalise, hash, cluster. Be honest about the ceiling: 33,514,469 addresses carry a postcode, only 423,458 carry coordinates (0.95%) and with_uprn_anchor is 0. This is a normalisation and matching service, not a geocoder Weeks

Monitoring and early warning · 9

Network and graph · 7

  • D51 Ownership Graph Product “Draw me the control structure.” 775,008 edges, 723,212 controlled companies, 260,079 ultimate owners, max depth 33, 1,232 in a cycle. Shipped as a resolved chain per company with depth, cycle flag and controller count Days
  • D54 Shared-Address Network “Which companies sit behind one door?” Components over the org–address bipartite graph; 32,289 addresses already classified mass_registration Days
  • D52 Cross-Border Presence Graph “Where else does this group operate?” Identity crosses a border only on a shared identifier, never on a name, so the graph is precise and deliberately narrow. Weeks
  • D53 Buyer–Supplier Bipartite Graph “Who trades with whom in public money?” 902,970 US suppliers × 304 buyers; 287,453 RO suppliers × 22,590 buyers; 31,499 GB suppliers × 7,687 buyers. Centrality, community detection, and dependence per supplier Weeks
  • D55 Common-Control Screen “Are these bidders really independent?” Sold as a graph query service, not a dataset, and the reason is measured. The bid-rigging screen returns one row against the August 2026 lake, and 92 if the group size is relaxed from three companies to two, because UK procurement records rarely carry a company number that reaches a the control register. Sell the query, priced per investigation Weeks
  • D56 Group Structure Resolver “Is this company part of a group, for size purposes?” Published entity-relationship records plus control-register edges, turning an unresolved group into a resolved consolidation. Weeks
  • D57 Officer and Controller Network “Who else is connected to this director?” Person-level graph. The hardest subject-axis position in the estate — it is about people by construction, so it ships as an aggregate, a flag or an investigation output, never as a file Months

Prediction · 8

  • D36 Financial Projection Service “What will next year's accounts look like?” Four declared tiers — trend, shrunk_delta, sector_median, refused — every row labelled with the tier that produced it; Shrinkage blends a firm's own growth with its sector's by how many deltas it has Engine built
  • D37 Predicted Digital Intensity Index “How digital is this firm, on the Eurostat scale?” 0–12 DII estimated from proxies because nobody publishes a per-firm value; never zero for a firm with no proxies, since zero is the *most* eligible a firm can be under a ceiling gate Engine built
  • D38 Growth and Hiring Signal “Who is expanding right now?” Visa filings, sponsor-licence additions, headcount deltas and research grants as leading indicators of headcount growth Weeks
  • D39 Late-Payment Prediction “Will this customer pay us on time?” Reported days-to-pay plus balance-sheet liquidity; the only product in the estate with a directly observed label for the thing being predicted Weeks
  • D34 UK Turnover Estimation “What does this company actually turn over?” The largest gap in the UK market, measured: 190,084,770 GB facts contain only 309,753 turnover values, and company_profile_open already carries a turnover_not_filed flag. Estimate from employees, total assets, sector and region, with an interval and a refusal tier Months
  • D35 12-Month Insolvency Probability “Which of these will fail?” Trained on 1,276,085 notices with the filing history that preceded them; published with a calibration curve and a backtest Months
  • D40 Site Closure Prediction “Which of these premises will close?” Inspection cadence, rating trajectory, occupancy churn at the address, operator distress Months
  • D41 Award-Win Probability “What are our odds on this tender?” Incumbency, buyer's historical concentration, competition level and category; scored per (buyer, category, bidder) Months

Scenarios and simulation · 9

  • D44 Max Fundable Ask “What is the largest grant we can request and still pass?” Derived by inverting the Scheme rather than hardcoding a ratio, and reports Not Binding where the pre-screen imposes no ceiling at all — a materially larger answer than the ratio, and the honest one Engine built
  • D42 Headroom Scenario “Does this deal pass if we drop the production machinery?” A named pinning of part of a project, under which the maximum achievable score is computed — it answers whether a deal passes before an application exists. Already a defined term in scoring/CONTEXT.md Weeks
  • D43 De Minimis Headroom “How much aid can this company still take?” What remains of the €300,000 ceiling over the rolling three-year window under Commission Regulation (EU) 2023/2831 — rolling, not three fiscal years, which was the superseded 1407/2013 and disagrees by up to a year precisely for the firms near the limit. Two of the input registers are internal-only; the output is one number per company Weeks
  • D46 Portfolio Stress Test “What happens to my counterparty book if turnover falls 15%?” Apply a declared shock to the ratio inputs and re-rank; delivered as a fragility ranking with the pre- and post-shock drivers, not as a single number Weeks
  • D49 Merger Concentration Test “What does this acquisition do to concentration in this market?” HHI before and after, with the combining parties resolved through the ownership graph rather than by name Weeks
  • D50 Licence Lapse Simulation “What capacity leaves this market when these licences expire?” Authorised vehicles, registrations and rated premises falling out by month under observed renewal rates Weeks
  • D45 Grant Call Demand Simulator “How oversubscribed will this call be?” The Slovak Recovery Plan register records, per call, the window, the money, the rules and the demand each one drew. Fit demand against those features and apply to an open call Months
  • D47 Procurement Entry Simulator “If we bid this category, who do we displace?” Incumbent share, buyer switching history, competition level; output is a ranked target list of (buyer, category) pairs with an expected displacement Months
  • D48 Supply Chain Shock Simulation “If this supplier fails, which of my buyers are exposed?” Remove a node from the buyer–supplier graph and propagate through ownership; exposure measured in awards and value at the buyer Months

Scoring and algorithms · 11

  • D16 EU SME Size Class Service “Is this company micro, small, medium or large under 2003/361/EC?” Sizeclass.py: headcount is the binding ceiling, then turnover or balance-sheet total; publishes Size Class Basis (standalone_declared / group_unresolved / group_unchecked) as a disclosure rather than a guess Engine built
  • D20 Landlord Distress Score “Which corporate landlords are in trouble?” Proven: ukcremap built 1,295,658 rows from the real lake and passed 70/70 gates. This is that product, sold as a score rather than a list Engine built
  • D19 CEE Arrears Score (SK / CZ) “Is this supplier behind with the state?” Three statutory arrears registers (tax, health insurance, social insurance) plus the Czech unreliable-VAT flag, weighted by recency and persistence; the social-insurance register is internal-only and contributes only to the score Days
  • D21 Address Risk Score “Is this counterparty's registered office a red flag?” Companies-at-address, premises class, churn of occupants, distance from trading address Days
  • D15 Grant Eligibility Engine (CZ / SK / HU) “Which of these 40,000 companies can apply to this call, and which will score well?” The engine exists. A call is a YAML file, not a release: Eligibility Gates with a Gate Reason in the customer's words, a points Pre-screen, a Borderline band (40–49 against a pass of 50, which "doubled the addressable list"), and Max Fundable Ask derived by inverting the pre-screen rather than hardcoding a ratio Weeks
  • D17 GRID Distress Score (UK) “Rank my 200,000-name book by probability of failure in 12 months.” Logistic model calibrated on 1,276,085 notices; 0–100 with named drivers; published as an Estimate in its own table with model id and version, never as a column inside the published layer Weeks
  • D18 Romanian Counterparty Score “The same, for Romania, where the income statement exists” As D17 plus margin and expense-ratio deterioration, which GB cannot support; supersedes the withdrawn sales-intel-11 Weeks
  • D22 Employer Risk Score “What is this employer's litigation and enforcement exposure?” Tribunal outcomes, safety-enforcement notices, pay-gap trajectory and sponsor-licence changes. One input carries individuals and contributes to the score only. Weeks
  • D23 ESG Integrity Score “Is this supplier's environmental story consistent with its record?” Claimed-versus-recorded: an ESOS notification or a modern-slavery statement against an enforcement action or a lapsed waste registration Weeks
  • D24 Federal Supplier Integrity Score “Should we onboard this US contractor?” Award concentration, exclusion history, safety rating, entity-age; scored, not listed Weeks
  • D25 Screening Signal Service “Does anything on this list touch a restricted party?” Matched through the spine rather than on raw strings, with score and evidence returned. Sold explicitly as a signal, never as compliance — trade_gov_csl's own screening_is_not_due_diligence rider travels with every response Weeks

How it works

Quote the ID. That is the whole process.

Every entry carries an ID — USFED, UKCREMAP — and that ID is the same code the build, the release manifests and our invoices use. Put it in the subject line at data@inference.institute and the enquiry lands against exactly one product with no clarifying round trip.

  • 01 Tell us the ID, what you would use it for, and whether it would be redistributed. The redistribution answer decides which upstream licences apply to you, and it is the question that most often changes the answer.
  • 02 We send the full sample pack for that ID — the real schema, a watermarked extract, the methodology note and the licence — so the file can be evaluated before anything is agreed.
  • 03 Terms and delivery are agreed in writing. Nothing about cost is published on this site, for any of our work; it is quoted against your use, not read off a page.

Not in the catalogue

Most of what we hold is not packaged, and some of what you need is not held yet.

The catalogue is the finished end of a much larger estate: over a hundred public sources under contract across the UK, the US and Europe, most of them onboarded for a specific piece of work rather than for sale. If the jurisdiction, the grain or the join you need is not on this page, it is worth asking — the answer is usually that it exists and is not packaged, occasionally that a publisher forbids it, and sometimes that it is a project rather than a file.

Data projects are the same conversation from the other end: a pipeline, a resolved entity spine, or a one-off extract built to answer a question you cannot currently ask. Those are engagements rather than catalogue entries, and they start the same way.

Request data we do not list → Discuss a data project →

Next step

Ask for the sample pack.

Every catalogue entry has one: the real schema, a watermarked extract and the licence, sent against the ID. It is the fastest way to find out whether a file answers your question.

Book a call → Email us →