The PSR's 88 percent problem

The Payment Systems Regulator's mandatory reimbursement regime went live on 7 October 2024. By year one, 88% of authorised push payment fraud losses were being refunded against the regulator's 66% baseline. £173m returned in twelve months. UK Finance's matching APP loss data corroborates the floor. On the headline measure, the policy did what it was designed to do.

The spread is the story. 88% is a sector average. The bank-by-bank range runs 60% to 98%. One bank reimburses six in ten. Another reimburses nineteen in twenty. The same fraud loss, the same statutory regime, two different consumer outcomes. The PSR's Q2 2026 thematic review will hit this directly. The Q3 cycle is where the divergence becomes a supervisory question.

Past 85% reimbursement, the internal economics flip. Below that level, prevention beats claims handling on cost. Above it, the marginal pound moves into claims. Fraud teams become reimbursement teams. The detection layer hollows out. The pattern Iris built was forecast in February 2024 and is now visible in the bank-level data the PSR has stopped publishing publicly.

The PSR was abolished on 11 March 2025 by HM Treasury, with functions absorbed into the FCA over an eighteen-month transition. The mandatory reimbursement regime survived the abolition. The supervisory cycle did not. The first FCA-led review of APP reimbursement performance lands in 2026 on a Consumer Duty footing, not a payment-systems one. The question changes with the lever.

The harder fraud went where the policy could not reach it. Romance fraud where the survivor will not testify. Coercion-driven authorised transfers that read on the payment surface as instructed. Investment scams stretched over months with consent visible at every transaction. The unreimbursed long tail of the £173m is not a residual. It is the class of harm the reimbursement test was never designed to detect.

Professor Karen Croxson at the FCA has been making the linked point on harm prevention since 2024: outcomes regulation works only if the regulated firm can see the harm in the data it already holds. Consumer Duty, in its mature 2026 form, lands the same question on Boards. The Non-Financial Misconduct rule lands it again on culture and senior management responsibilities. Three rules, one question. Could you have seen this earlier?

The bit nobody has flagged is that 88% means the next regulatory cycle is not "increase reimbursement". It is "the bank that did not see it was not trying hard enough". The FCA's s166 template is being written now on that thesis. The first published case will not be a complaint failure. It will be a Board failure to act on a pattern visible in its own data.

This is the architecture question on the FS surface. Online platforms have been asked it for eighteen months. The Payment Services Regulator answered it for the banking sector with a reimbursement floor. The FCA's job, post-abolition, is to make the answer hold past the point where reimbursement stops working.

The two-week Meta story in May runs the same logic at a higher altitude. Architect away the audit trail, then claim you could not have known. The PSR's 88% closed the version of that argument that was available to the banks. The next version was already moving when the floor was set.

FS Boards reading the year-one data should be asking the question the PSR could not answer in its own register. At what reimbursement level does the firm stop trying to see the harm at all? The Consumer Duty answer is now: never. The s166 enforcement answer arrives later this year.

Iris is built for the architecture answer. Pattern visible at week three, not week eleven. The reimbursement clock and the detection clock are different clocks.

Sources: PSR mandatory reimbursement regime, effective 7 October 2024; PSR abolition by HM Treasury, 11 March 2025; FCA absorption of PSR functions, transition through 2026; UK Finance Annual Fraud Report 2025-26; FCA Consumer Duty (in force July 2023, mature application 2026); FCA Non-Financial Misconduct rule (in force 1 September 2026); Professor Karen Croxson, FCA Chief Economist.

← Back to Iris · caroline@trustiris.com

Confidential. (c) 2026 Iris Anticipa Ltd. All rights reserved.