United KingdomPublic sector2025 – ongoingDepartment for Work and Pensions (DWP)
Universal Credit Advances fraud-risk model
DWP says its model for Universal Credit advances finds fraud risk 2.5 times better than random checks
Problem
Universal Credit advance payments are made quickly, and DWP estimates fraud and error on them at between £20 million and £90 million in 2025 to 2026.
Approach
A supervised classifier scores each advance request before payment and sends high-risk ones to a human reviewer. From April 2025 to March 2026, DWP gave 1.3 million advances worth about £700 million. DWP has retrained the model but says it needs more testing before the new version goes live.
Results
Takeaway
Flagged requests approved by a person were paid a median of one day later than automatic approvals. DWP gives the effect size as a comparison with a random sample, and withholds how often the model runs.
Categories
- Size
- Large enterprise
- Industry
- Public sector
- Business function
- Finance, risk and legalProduction and quality
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and ML
- Adoption
- Built in-house
- Evidence
- Company's own claim
Sources
Source summaries
DWP describes the model, its human review step and its measured effectiveness against a random sample. It also gives the size of the advances programme and fraud estimates.
Full details are in the original sources.
