HMRC has been urged to strengthen governance and risk management after a data-matching exercise aimed at tackling Child Benefit fraud and error led to thousands of eligible claimants having payments stopped without warning.

A National Audit Office (NAO) report examined HMRC’s use of Home Office travel data to identify claimants who may have been incorrectly receiving Child Benefit while living outside the UK.

The watchdog said the first rollout generated estimated fraud and error savings of £60m, but more than 8,000 genuine claimants had payments suspended. By April 2026, HMRC information showed 59% of the initially suspended claims had been confirmed as eligible.

HMRC began developing the approach after identifying in 2021 that Child Benefit records could be matched with airline data to flag people absent from the UK for more than 12 weeks. A 2024 pilot prevented an estimated £15m in incorrect payments, prompting a wider rollout in August 2025.

However, the NAO found that HMRC changed the process between pilot and rollout, including removing an early Pay As You Earn (PAYE) check. This led to a much higher number of eligible claimants being caught up in the exercise. HMRC also sent longer and more complex letters and questionnaires asking claimants to prove eligibility.

By October 2025, HMRC had suspended payments for 23,794 claimants. Some reported financial and emotional impacts from the sudden loss of income, as well as stress and difficulty proving their entitlement. Between August 2025 and February 2026, HMRC handled more than 22,500 calls linked to the rollout.

HMRC took corrective action once problems emerged in mid-October, including reintroducing a PAYE check, automatically reinstating some payments and allowing claimants to confirm eligibility by telephone. It also apologised to affected claimants and, by April 2026, had paid £3,200 compensation to 51 people.

The NAO said HMRC accepted there were weaknesses in the transition from pilot to rollout, including shortcomings in risk assessment and decision-making. Key decisions were made without sufficient scrutiny, and the department did not adequately assess the impact on claimants whose payments were wrongly stopped.

HMRC relaunched the initiative in March 2026 with changes, including giving claimants one month to prove eligibility before payments are suspended and enabling evidence to be submitted online. It has also commissioned an internal audit and is taking what the NAO described as a more controlled approach.

The department projected gross savings of £366m over the five years to 2029-30, based on the pilot, and plans to update its cost and benefit estimates for the Autumn Budget 2026.

The NAO recommended that HMRC clarify its risk appetite for future decisions, improve performance data so it can monitor claimant experience alongside compliance outcomes, and share lessons across government fraud and error teams.

Gareth Davies, head of the NAO, said it was right for HMRC to seek new ways to tackle fraud and error, and that the initial work suggested significant savings could be secured for taxpayers. He said: “However, missteps in implementing the first rollout meant HMRC did not strike the right balance between detecting fraud and error and managing the impacts on Child Benefit claimants. While HMRC should not be discouraged from pursuing innovative ways to reduce fraud and error, it must learn and apply the lessons for future initiatives.”