FrancePublic sector2024 – ongoingCNAF
Predictive risk scoring to select benefit checks
CNAF reported 166,000 benefit checks selected through predictive scoring in 2024 amid a legal challenge
Problem
French family benefit payments depended partly on declarations that could contain errors. CNAF needed to select files for investigation with limited inspection staff.
Approach
CNAF used a data-mining model to rank benefit files by the risk of overpayment. Its technical documentation described logistic regression and said inspectors considered scores alongside other information. Advocacy groups challenged the scoring system in October 2024; CNAF subsequently documented a revised model for 2026.
Results
Takeaway
CNAF described scores as decision support for inspectors. The legal challenge concerned the scale of profiling and alleged discrimination, rather than a demonstrated productivity gain.
Categories
- Size
- Large enterprise
- Industry
- Public sector
- Business function
- Finance, risk and legal
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and ML
- Adoption
- Built in-house
- Evidence
- Company's own claim
Sources
Source summaries
CNAF explained the statistical model used to help inspectors select benefit files. It reported 166,000 model-originated checks in 2024 and described the revised system and human selection process.
A claimant organisation explained the coalition's challenge to CAF scoring before the Conseil d'État. It alleged discriminatory effects and excessive profiling; these were the claimants' arguments rather than court findings.
Full details are in the original sources.