The UK's DWP fraud model: a machine-learning system that saved an estimated 4.4 million pounds on Universal Credit advances
Since May 2022 the UK Department for Work and Pensions has run a machine-learning model that flags potentially fraudulent Universal Credit advance claims for review. Per the National Audit Office, the model saved an estimated 4.4 million pounds and was around three times more effective at identifying fraud risk than a randomised control group. The figures come from DWP and are relayed by the NAO, the UK's independent public auditor; the Public Accounts Committee has raised concerns.
| Metric | Before | After |
|---|---|---|
| Estimated savings from the fraud model | about 4.4 million pounds | |
| Effectiveness at identifying fraud risk | randomised control group sample | around three times more effective |
| In production since | May 2022 |
The problem
Universal Credit advances are paid quickly, which is exactly what makes them a target for fraudulent claims. Screening every advance by hand is not feasible at the scale the Department for Work and Pensions operates, so DWP turned to data analytics to decide which claims warrant a closer look.
What was built
Per the National Audit Office, “since May 2022, DWP has used a machine learning model to flag potentially fraudulent claims for Universal Credit advances” (source). The model scores claims for fraud risk and routes the flagged ones to human reviewers rather than deciding outcomes on its own.
The outcome
An estimated 4.4 million pounds saved. Per the NAO, the model has been “saving an estimated £4.4 million” (source).
Around three times more effective than random sampling. “DWP found the machine learning model to be around three times more effective at identifying fraud risk than a randomised control group sample” (source). The comparison against a randomised control group is what makes this figure more than an assertion, though the underlying numbers are still DWP’s own estimates.
The counter-context. The same NAO release notes that “concerns have been raised by the Public Accounts Committee” about DWP’s wider fraud and error position (source). This file reports the model’s stated results and that caveat together.
How this was verified
This case is pending. Its footing is better than a vendor case study: the figures reach the public through the National Audit Office, the UK’s independent public-spending auditor, and rest on a comparison against a randomised control group. But they are still DWP estimates, single-sourced, and not independently re-measured, and the Public Accounts Committee has flagged concerns about the broader picture. Green requires an independent measurement or DWP confirming the savings estimate and its basis on the record.
Sources
Cited in this case file. Tier 1 = independent public auditor relaying government figures. Each figure was checked against the live source on 2026-07-10.
- National Audit Office, “DWP begins to make headway tackling benefit fraud and error” (Tier 1, independent auditor; figures are DWP estimates). https://www.nao.org.uk/press-releases/dwp-begins-to-make-headway-tackling-benefit-fraud-and-error/ · archived
In-house machine-learning model flagging Universal Credit advance claims for reviewHuman review of flagged claims
- Status
- pending
- Method
- Sourced by The Internet Ninja against the public record: every figure below is quoted from the cited source and, on 2026-07-10, checked against the live page over the open network, with a web.archive.org capture taken the same day. This case is pending, it has not yet cleared TIN's green validation bar. The source is the National Audit Office, the UK's independent public-spending auditor, which relays figures produced by DWP; the numbers are DWP estimates reported by an independent body rather than DWP marketing, which is a stronger footing than a vendor case study, but they remain single-sourced and DWP-originated. The NAO also notes the Public Accounts Committee has raised concerns. A green badge requires an independent measurement or DWP confirming the savings estimate and its basis on the record.
- Provider
- UK Department for Work and Pensions, in-house machine-learning model
- Client
- UK Department for Work and Pensions (DWP) · Government / social security (UK)
- Disclosure
- named
What did the DWP machine-learning fraud model do?
Since May 2022 it has flagged potentially fraudulent Universal Credit advance claims for human review. Per the National Audit Office, it saved an estimated 4.4 million pounds and was around three times more effective at identifying fraud risk than a randomised control group.
Why is this case file pending, not verified?
The figures are DWP estimates relayed by the National Audit Office; they are single-sourced and DWP-originated, and the Public Accounts Committee has raised concerns. A green badge requires independent measurement or DWP confirming the estimate and its basis on the record.