System 17 · Free interface, structured identifiers, unused

Watchman

Insolvency exposure across public suppliers

Three million insolvency notices, free, each carrying a company number. Nobody joins them to the register of who holds public contracts.

Fraud & integrityPublic moneyOperations

The problem

When a company holding public contracts fails, the consequences land on whoever depended on it. The failure of a major government contractor in 2018 remains the defining example, and the lesson drawn afterwards was that nobody had a consolidated view of exposure.

The raw material to build that view is free and better than expected. The official insolvency record publishes a JSON interface with no key required, carrying over three million notices. Critically, each notice carries a structured company number — a direct identifier join, with no name matching required.

On the other side sit the public procurement record, the care provider registers, the school trust register and the social housing register — millions of contractual and regulatory relationships between the state and companies.

Nobody joins them. So the question "which public contracts and regulated services are currently held by companies in insolvency proceedings" has no answer, despite both halves being free and one of them carrying a clean identifier.

And the standard alternative does not work. Predicting failure from filed accounts requires profit and loss data that small companies do not publish — a requirement that was paused in January 2026 and pushed to 2028, with an opt-out from publication attached. Roughly two-thirds of the standard model’s power is unavailable, and published testing shows such models flagging around a quarter of all companies, of which the overwhelming majority never fail.

3m+
Insolvency notices available free, each carrying a company number
0
Systems joining them to the register of public contracts
2028
Earliest date small company accounts might carry profit and loss — with an opt-out attached
~1 in 198
Annual company failure rate, against which any predictive model must justify itself

The system

Watchman abandons prediction and does something more useful: it watches events, and it knows who depends on whom.

It ingests insolvency notices continuously, resolves each to an organisation through the entity spine, and checks that organisation against every public relationship the platform knows about — contracts, care registrations, school trusts, social housing, land holdings.

The output is an exposure alert rather than a risk score. Not "this company might fail" but "this company has entered administration, and here are the seventeen contracts and four hundred care beds that depend on it."

It layers earlier signals underneath — charge filings, strike-off action, overdue accounts — all observable for every company regardless of size, and all events rather than inferences.

And it inverts for planning: for any council or department, which of its suppliers show distress signals, and what would be lost if each failed.


Worked examples

Two situations this system answers

EXAMPLE 1
The problem

A supplier enters administration on a Friday afternoon. Which departments, councils and services depend on it? Nobody can say before Monday, and possibly not for weeks.

What Watchman does

Watchman reads the notice. Where it carries a company number — every liquidator appointment does — the public bodies relying on that company are listed immediately, with no guessing about which company it is. Where the number is missing, the registered address in the notice still places it.

EXAMPLE 2
The problem

A small supplier files accounts late two years running and takes on new charges over its assets. Both are published signals of difficulty. Nobody is watching, because the standard financial models need profit data small companies never file.

What Watchman does

Watchman treats those as facts that have happened rather than a prediction, and warns the buyers who depend on that supplier.


Data foundation

Every dataset below is open, or its access constraint is stated

DatasetPublisherWhat it provides
Insolvency noticesThe GazetteFree JSON interface, no key, over three million notices, each with a structured company number and a direct link to the company record.
Companies House bulk and streamsCompanies HouseCompany data, charges, insolvency and control. Free bulk downloads plus resumable event streams.
Procurement recordCabinet OfficeContract awards and performance across the public sector.
Care provider registersCQC and OfstedRegulated services and their operators.
Academy trust membershipDfECompany numbers on every open trust.
Registered providersRegulator of Social HousingSocial housing providers, with a corporate form field routing each to the right identifier authority.
Local authority spendingCouncilsPayments to suppliers, once normalised and resolved.

Capabilities

01
Event ingestion
Insolvency notices processed continuously, resolved to organisations by identifier rather than by name.
02
Exposure alerting
When an organisation enters proceedings, everything the public sector depends on it for, immediately.
03
Early signals
Charge filings, strike-off action and overdue accounts — observable for every company, including those filing minimal accounts.
04
Buyer-side view
For any council or department, which suppliers show distress and what each failure would cost.
05
Concentration risk
Where many public bodies depend on one organisation, or one group behind several apparent suppliers.
06
Group awareness
Distress in a parent or sibling company that the contracting entity’s own record would not show.
07
Historical validation
Back-tested against known failures, with performance published rather than claimed.

Benefits

For government

  • Gives the public sector the consolidated supplier exposure view whose absence was the lesson of the 2018 contractor collapse.
  • Works for small suppliers, where accounts-based models cannot, because it uses events rather than ratios.
  • Supports the care market oversight duties that now sit in statute, from the supplier side.
  • Lets buyers act before a failure rather than discovering exposure afterwards.

For the public

  • Services less likely to stop abruptly when a provider fails.
  • Public money less likely to be committed to organisations already in difficulty.
  • Transparency over which public services depend on which private companies.

Delivery

Likely sponsor
Cabinet Office commercial, Public Sector Fraud Authority, CQC and Ofsted oversight, individual councils
Procurement route
SBRI for the resolution and alerting engine, then G-Cloud for the service

Phasing

1
Ingest and resolve
3 months
Insolvency notices flowing and resolving to organisations by identifier.
2
Exposure joins
4 months
Connected to procurement and provider registers across the platform.
3
Early signals
3 months
Charges, strike-off and filing delinquency added as leading indicators.
4
Buyer tooling
5 months
Council and department supplier risk views released.

Risks & mitigations

Do not build a failure predictor

The base rate is roughly one in two hundred companies a year. Any model flagging a meaningful share will be wrong the overwhelming majority of the time. Watchman reports events that have happened and dependencies that exist, not probabilities of future failure.

Reputational harm

Publishing that a company shows distress signals could itself precipitate difficulty. Exposure analysis stays in a restricted tier for public bodies; only aggregate concentration is published.

Identifier gaps

Several provider registers carry no company number, and under half of social housing providers are Companies House entities at all. Coverage is reported per register rather than assumed uniform.


Sources

All sources checked in August 2026. Figures carry the reference period of their source, which may differ from publication date. Where a figure could not be verified against a primary source it is not used.