UK payment firms should treat APP scam reimbursement as a product and operating-model obligation, not a compliance afterthought. The payment journey, evidence record, claims queue, sending-side and receiving-side controls, customer support and loss economics must work as one system, with clear ownership from payment initiation through reimbursement and recovery.
- Key takeawaysMap eligibility and ownership across every payment rail and access partner.
- Decision frameworkA UK payment firm should design APP scam reimbursement into the payment journey, loss model and support operation.
- Start with the payment and customer in scopeThe PSR set the Faster Payments reimbursement cap at £85,000 per claim from 7 October 2024.
- Measure the cost per legitimate paymentThe useful commercial measure includes more than the fraud written off.
- Put friction where it can change a decisionA generic warning becomes easy to dismiss.
Answer first: UK payment firms should treat APP scam reimbursement as a product and operating-model obligation, not a compliance afterthought. The payment journey, evidence record, claims queue, sending-side and receiving-side controls, customer support and loss economics must work as one system, with clear ownership from payment initiation through reimbursement and recovery.
Key takeaways
- Map eligibility and ownership across every payment rail and access partner.
- Measure customer disruption and claims operations alongside fraud loss.
- Preserve the exact warning, response and transaction evidence seen in each case.
- Review receiving-account acquisition and sending controls as one risk system.
Decision framework
| Decision area | Evidence to require | Risk it controls |
|---|---|---|
| Scope | Eligible customer, rail, date, timing and exclusions are machine-readable | Misrouted or delayed claims |
| Intervention | Warning responds to payment context and is versioned | Generic friction without learning |
| Claims | Queue owner, evidence request and clock status are visible | Silent ageing and complaints |
| Economics | Loss, recovery, support and false-positive cost share one view | Underpriced payments |
| Receiving risk | Acquisition source and post-onboarding activity are connected | Repeat mule-account exposure |
A UK payment firm should design APP scam reimbursement into the payment journey, loss model and support operation. A warning screen alone cannot carry the obligation. The firm needs to recognise eligible claims, preserve transaction evidence, communicate clearly and finance losses without treating legitimate customers as suspects.
Authorised push payment fraud occurs when someone is deceived into authorising a payment. That matters because the instruction may look technically valid. A product that checks only whether the correct customer approved a transfer can still miss the reason the customer approved it.
Start with the payment and customer in scope
The PSR set the Faster Payments reimbursement cap at £85,000 per claim from 7 October 2024. Its consumer guidance also explains protection for qualifying UK transfers through Faster Payments and CHAPS, subject to the applicable rules. 12 These are not universal protections for every payment, customer or commercial dispute.
Build a scope table covering customer eligibility, payment rail, transaction date, claim timing, exclusions and vulnerability. Keep CHAPS and Faster Payments rule sources distinct even where their outcomes align. A customer-service agent should not have to infer the scheme from a payment's appearance in an app.
For a London fintech using another institution's payment access, document who receives the claim, who investigates, who communicates and who funds reimbursement. Commercial allocation between firms should not become an invisible gap in the customer's route to help.
Measure the cost per legitimate payment
The useful commercial measure includes more than the fraud written off. Add investigation time, customer contact, recovery activity, complaints and the cost of delaying legitimate transfers. Separate sending-side exposure from receiving-side exposure when the relevant arrangements require it.
Consider an illustrative product earning £0.40 contribution on each of 100,000 monthly transfers before fraud and claims operations. That creates £40,000. A £12,000 monthly claims-and-investigation cost consumes 30% of that amount. This is arithmetic, not an estimate of industry performance. It shows why a pricing discussion cannot ignore the loss process.
Track the distribution as well as the average. Several quiet months can conceal a business model that cannot absorb one concentrated scam event. Finance and risk teams should agree stress assumptions before marketing accelerates acquisition.
Put friction where it can change a decision
A generic warning becomes easy to dismiss. A useful intervention responds to the payment's context: a new beneficiary, an unusual amount or a customer explaining that an apparent authority instructed the transfer.
Test whether the intervention changes behaviour. Record abandonment, successful legitimate payments, subsequent scam reports and requests for human assistance. Do not describe a higher warning-display rate as a better customer outcome.
There is a practical tension. Too little friction exposes customers to preventable harm; indiscriminate blocking can prevent rent, payroll or urgent supplier payments. Product reviews should examine both sides using defined cohorts, rather than celebrating a falling fraud-loss number in isolation.
Preserve the decision trail
Store the warning version, information presented, customer response, relevant timestamps and subsequent actions. Access should be limited and retention justified. A screenshot of the current interface cannot establish what a customer saw three months earlier.
The claim record should distinguish verified facts, customer reports and investigator judgement. That separation makes internal review more useful and reduces the temptation to turn an uncertain inference into a refusal reason.
Give support staff a route to escalate vulnerability and unusual circumstances. The PSR's guidance contains conditions and timing requirements; implementation should translate these into monitored queues and clear ownership. 2
Use external comparisons carefully
The PSR's dashboard defines its reporting population and periods. It also explains differences from other fraud datasets. 3 Comparing a firm's internal loss rate with a public industry number is unhelpful unless payment types, denominators and claim maturity align.
A practical monthly review should show eligible claims received, decision time, reimbursement, recoveries, overturned decisions and customer complaints. Add a sample of completed cases, including legitimate payments disrupted by controls. Numbers identify where to look; case review reveals what the process actually did.
Review the receiving side as a product risk
An attractive account-opening journey can also attract accounts used to receive scam proceeds. Review the connection between acquisition, account activity and investigation. Do not let growth reporting treat every opened account as an equally desirable outcome.
Consider a hypothetical increase in new accounts followed by several unrelated senders reporting fraud. The investigation should establish the facts and relevant obligations, but the product review should also ask which acquisition channel, verification step or account feature created exposure. Closing individual cases without examining that pattern can leave the source untouched.
Keep measures for legitimate new customers alongside the risk indicators. Tightening every onboarding step may reduce exposure while excluding suitable businesses. A targeted change needs a defined population, expected effect and review date.
Allocate the cost of uncertainty
Decide how unresolved claims enter management reporting. A case awaiting evidence should not be counted as a successful rejection merely because cash has not yet left the firm.
Finance can show estimated exposure separately from confirmed expense, with assumptions visible.
Compare the early estimate with later outcomes. If losses repeatedly arrive above the estimate, investigate whether the problem is missing cases, optimistic recovery assumptions or slow recognition. This feedback makes the product's contribution calculation more credible and helps management fund support before a queue becomes a customer problem.
What should change first?
Map one payment from initiation to a possible reimbursement claim. If ownership becomes unclear at any step, fix that handover before buying another fraud tool. The strongest control is a connected operation in which product, risk, finance and support can explain the same transaction.
A useful monthly review puts claims, losses, disrupted legitimate payments and review costs on the same page. That is where a payment firm can see whether its product is becoming safer to use and commercially stronger. Scheme eligibility still depends on the relevant rules and the facts of the claim.
A practical 90-day implementation plan
- Map one Faster Payments claim end to end, including every partner hand-off.
- Create a common claims-and-control data dictionary for product, risk, support and finance.
- Review a sample of reimbursed, rejected and interrupted legitimate payments.
- Approve a monthly scorecard with decision times, losses, recoveries, complaints and customer disruption.
Frequently asked questions
Does Confirmation of Payee remove APP reimbursement risk?
No. It can reduce some misdirected payments and some fraud patterns, but Pay.UK expressly says it is not a guarantee that fraud will be detected or reimbursed.
What should a product team retain?
Retain the warning version, customer response, timestamps, payment context, subsequent action and an auditable distinction between verified fact and investigator judgement.
Should a firm optimise only for lower fraud loss?
No. A control can reduce fraud while creating unacceptable disruption for legitimate payments. Review both outcomes by customer and payment cohort.
Reporting by Shoreditch Talk




