APP Fraud Reimbursement in the UK: The Fraud Stack, Costs and Timeline PSPs Need
UK reimbursement rules put part of every APP scam claim on both the sending and receiving PSP. Here's the fraud stack, indicative cost ranges and a 9–12 month rollout plan to limit what you pay out.
Why reimbursement liability changes what you buy
If you run fraud or payments at a UK bank, e-money institution (EMI) or payment institution that sends or receives Faster Payments, mandatory reimbursement has turned authorised push payment (APP) fraud losses into a balance-sheet line. Before the regime began in October 2024, most losses landed on victims, or were covered by a voluntary code that only some firms signed. Now reimbursement is mandatory for in-scope consumer, micro-enterprise and charity payments, and the cost is split 50/50 by default between the sending and receiving payment service provider (PSP).
That split should drive how you shop for APP fraud prevention software UK firms rely on. A sending bank can no longer treat inbound mule activity as someone else's problem. A small EMI that mostly receives funds now pays for every mule account it onboards. The old question was "how do we stop our customers being scammed?" There are now two: how do we stop risky outbound payments, and how do we avoid being the account that receives the money?
This is a UK country guide. The Payment Systems Regulator (PSR) set the rules for Faster Payments, and the Bank of England covers CHAPS. The government has announced that the PSR's functions will move into the FCA, so check the current rulebook and take compliance advice before you finalise controls. Firms elsewhere should still watch this space: Australia's Scams Prevention Framework, Singapore's Shared Responsibility Framework and proposed EU payment services reforms all push liability toward PSPs in similar ways.
The reimbursement parameters that shape your controls
The headline rules are short, but each one creates a tooling or process requirement. The figures below reflect the regime as we understand it. Confirm them against current PSR, FCA and Pay.UK publications, because the cap and guidance have been revised before.
| Rule | How it generally works | What it demands from your stack |
|---|---|---|
| Liability split | 50% sending PSP, 50% receiving PSP by default | Inbound mule detection becomes a direct loss control |
| Maximum claim | A per-claim cap in the tens of thousands of pounds (currently £85,000) | Prioritize high-value outbound scoring and payment holds |
| Reimbursement deadline | Five business days, with limited "stop the clock" for evidence gathering | Case management that tracks SLA timers per claim |
| Claim window | Roughly 13 months after the final payment | Data retention and evidence retrieval across that period |
| Exceptions | Gross negligence (consumer standard of caution) or first-party fraud; vulnerable customers get extra protection | Documented warnings, interventions and vulnerability flags |
| Optional excess | Sending PSP may apply a small excess (about £100); not for vulnerable customers | Policy configuration in the claims tool |
Two related changes matter too. Amendments to the Payment Services Regulations let PSPs delay an outbound payment for a limited number of business days where they have reasonable grounds to suspect fraud. Confirmation of Payee now applies to a much wider group of PSPs. Both give you more places to intervene, but only if your scoring is accurate enough to justify the friction.
Outbound: payee risk scoring beyond name checks
Confirmation of Payee (CoP) catches misdirected payments and crude impersonation. It misses the most expensive cases, such as investment and romance scams, where victims pay a correctly named account they've been coached to trust. For those you need payee risk scoring: a real-time assessment of the destination account, not only the sender's behaviour.
Strong outbound tooling combines three signals:
- Payee intelligence: the age of the receiving account, how many unrelated senders have paid it recently, and whether it's been reported. Mastercard's Consumer Fraud Risk, built on Vocalink payments data, is the best-known network-level option in the UK.
- Behavioural and session signals: hesitation, remote-access tools, an active phone call during a banking session, and pasted payee details. BioCatch is the established name here. Featurespace, Feedzai and NICE Actimize offer comparable modules.
- Customer context: a first payment to a new payee, payment size against balance and history, and vulnerability flags.
Our view: most mid-sized banks should buy network payee intelligence before adding another behavioural layer. Behavioural tools are good at spotting a customer under coaching. Payee data can tell you that 30 strangers paid the same account this week, and that signal is much harder for scammers to defeat.
Inbound: mule account detection as a liability control
Under the 50/50 split, the receiving PSP funds half of every reimbursed claim that lands in its accounts. Neobanks and EMIs with fast digital onboarding tend to be disproportionately exposed, because mule recruiters target them for exactly that reason.
Detection works at two points. At onboarding, look for synthetic or borrowed identities, device reuse, and applicant profiles that match known recruitment patterns, such as students or recent arrivals. Once the account is live, monitor inbound flows: many unrelated payers, money in and out within minutes, and activity that doesn't fit the stated purpose. Vendors include Featurespace, Feedzai, Hawk, Sardine and LexisNexis Risk Solutions. Network-level flow tracing from scheme operators adds visibility no single institution can replicate.
Hypothetical example: say an EMI is the receiving PSP on 60 reimbursed claims a year averaging £4,000 each. Its 50% share is about £120,000 a year before handling costs. An inbound monitoring tool costing the equivalent of $75,000–$190,000 a year only pays for itself if it cuts mule inflows sharply, and you can only prove that if you can measure claims per account cohort. So the claims data matters as much as the detection model. At this volume, the case for a premium tool is thin. Start with consortium data and onboarding controls.
Data-sharing consortia: where most of the signal sits
No single PSP sees a full scam journey. The money crosses institutions, and first contact usually happens on a social platform, marketplace or phone call. Consortium data closes part of that gap:
- Cifas National Fraud Database membership for confirmed mule and identity-fraud markers.
- Network-level services from scheme operators, which score accounts using payment flows across many PSPs.
- Sector intelligence sharing through UK Finance programmes, and cross-sector work with telecoms and tech platforms through Stop Scams UK.
Before you join, check the data-protection basis for sharing, retention periods, and whether your vendor can ingest consortium data natively or needs custom integration. UK GDPR and recent data-protection reforms both affect how far you can share and act on these signals. Get your DPO and legal counsel to sign off on the processing basis.
Claims workflow: the component most firms underbuild
Many firms spend heavily on detection, then run claims on spreadsheets and shared inboxes. With a five-business-day deadline and a counterparty PSP that must settle its share, that fails. Directly connected Faster Payments participants generally use Pay.UK's Reimbursement Claims Management System (RCMS) to exchange claims and recover the other half. Check Pay.UK's current requirements for indirect participants.
A workable claims stack needs:
- Intake that captures scam type, payment chain and vulnerability indicators at first contact
- Automatic retrieval of the warnings, interventions and session evidence shown before each payment, which you need to apply any standard-of-caution exception
- SLA timers for reimbursement, stop-the-clock events and inter-PSP settlement
- Reporting feeds for regulatory returns and internal loss analytics by channel, payee type and customer cohort
Case management modules from NICE Actimize, SAS, Feedzai and others cover much of this. Smaller firms often configure general workflow platforms instead. Either can work. Before you commit, make the vendor produce a defensible evidence pack for a single claim in minutes, live, on your data.
Indicative costs by firm profile
The ranges below are typical 2026 market ranges in USD for annual software and data costs. Local pricing differs, and contracts vary widely with transaction volume and bundling, so confirm with vendors. Implementation commonly adds roughly 50–150% of first-year licence costs.
| Firm profile | Outbound scoring and CoP | Inbound mule detection | Consortium and network data | Claims workflow |
|---|---|---|---|---|
| Small EMI or payment institution (under 250k accounts) | $40k–$150k | $50k–$150k | $15k–$60k | $20k–$80k |
| Mid-sized bank or neobank | $150k–$600k | $150k–$500k | $60k–$250k | $80k–$300k |
| Large retail bank | $1m–$4m+ | $600k–$2m+ | $250k–$1m+ | $300k–$1.2m+ |
The main cost drivers are scored transaction volume, the number of channels instrumented (app, web, open banking), per-query network data fees, and whether you buy a platform or point solutions. Budget for fraud operations staff as well. A model that raises alerts nobody reviews inside the payment-delay window doesn't reduce losses.
A realistic rollout timeline
For a firm starting from basic rules-based monitoring, plan on 9–12 months. Firms with existing platforms can compress that.
| Phase | Months | Key deliverables |
|---|---|---|
| Exposure baseline | 0–2 | Claims by scam type, sending vs receiving split, top mule cohorts, gap assessment against current rules |
| Claims workflow and RCMS readiness | 1–4 | SLA tracking, evidence packs, inter-PSP settlement process |
| Inbound mule detection | 2–7 | Onboarding controls, flow monitoring, consortium data feeds |
| Outbound payee risk scoring | 3–9 | Network payee intelligence, dynamic warnings, payment-delay policy |
| Tuning and governance | 6–12 | False-positive review, model validation, board reporting |
Put the claims workflow first. It's the cheapest phase, it cuts regulatory risk immediately, and it produces the loss data you need to choose and tune everything after it.
Red flags and a decision rule for vendor selection
Treat these as warning signs when vendors pitch you APP fraud prevention software UK buyers are being offered:
- Detection rates quoted without the false-positive rate or the customer friction that comes with them
- No native support for both outbound and inbound use cases, which leaves you integrating two vendors to cover the 50/50 split
- Network or consortium data described vaguely, with no clarity on sources, refresh frequency or data-protection basis
- Claims functionality that's only on the roadmap, or can't export evidence for a disputed standard-of-caution decision
- Pricing per alert rather than per scored transaction, which charges you more as your models get noisier
The decision rule: if more than half your reimbursed value comes from the receiving side, fund mule detection and consortium data before upgrading outbound scoring. If most of it comes from the sending side, buy network payee intelligence first. Either way, ask shortlisted vendors to run a retrospective test on 6–12 months of your own claims data before you sign, and have compliance counsel review your exception and vulnerability policies against the current rules.
Common questions
Does the APP fraud reimbursement requirement apply to EMIs and indirect Faster Payments participants?
Generally yes. The PSR's requirement applies to PSPs that provide relevant accounts and send or receive Faster Payments, including indirectly connected firms. Reporting and claims-exchange arrangements can differ for indirect participants, so check current Pay.UK and regulator guidance and confirm your position with compliance advisers.
Is Confirmation of Payee enough to meet our obligations?
No. It helps with misdirected payments and some impersonation scams, but it doesn't stop victims who knowingly pay a correctly named account. You still need payee risk scoring, effective warnings and, as a receiving PSP, mule detection to reduce reimbursed losses.
Can we refuse reimbursement if the customer ignored our warnings?
Only in limited cases. The consumer standard of caution exception generally requires gross negligence, which is a high bar, and it doesn't apply to vulnerable customers. You'll need well-documented, specific warnings and interventions to rely on it, so take legal advice on your policy.
How long does it take to get a mule detection tool live?
For a small or mid-sized firm, basic inbound monitoring and consortium feeds commonly take three to six months, including data integration and initial tuning. Onboarding controls and full model tuning usually take longer. A retrospective test on your own claims data shortens tuning.
Are these rules relevant if we operate outside the UK?
If you have UK customers or a UK-licensed entity, they apply directly. Elsewhere, similar shared-liability approaches are emerging, for example in Australia, Singapore and proposed EU reforms. Building outbound and inbound controls together now makes it easier to adapt as those regimes develop.
Written by the Equitystack Pro editorial team. Prices, features and rules change often - confirm details with the vendor or a licensed professional before you decide.