Join our Newsletter — 33% off our NHI Course
Home› Guides› Fraud Prevention Maturity Model: Five Levels to Benchmark…
Guide Fraud Prevention

Fraud Prevention Maturity Model: Five Levels to Benchmark Your Fraud Programme

PDF for newsletter subscribers · free to read online

← All guides
By NHI Mgmt Group, edited by Lalit Choda Updated 27 October 2026 13 min read
On this page

Most organisations can tell you their fraud losses last quarter, but few can say how mature their fraud prevention capability is or what would move it forward. Fraud budgets tend to follow the last big loss: a rules engine after a wave of card testing, a verification vendor after an identity fraud spike. This guide sets out NHI Mgmt Group's fraud prevention maturity model, which uses our standard five levels, Ad hoc, Developing, Defined, Managed and Optimised, and scores them across nine capability areas. The model is our own; it draws on public frameworks such as the GAO Fraud Risk Framework and the UK government guidance on the failure to prevent fraud offence, which are cited where used. It is for fraud, financial crime and risk leads who need to benchmark their programme, agree a target with the board and plan the next 12 months. For how to build the programme itself, see the Fraud Prevention Guide for Digital Businesses.

Key takeaways

  • Five levels, nine capability areas. Score each area separately; the overall level is the lowest area that matters for your top fraud risks, not an average.
  • Level 3 Defined is the first realistic target for most firms: a named owner, a written fraud risk assessment, documented controls for each priority fraud type, and loss and friction metrics reported monthly.
  • Two kinds of fraud, one programme. Fraud against the organisation and its customers, and fraud committed by people acting for the organisation, now both need evidence of reasonable procedures in the UK since the failure to prevent fraud offence came into force on 1 September 2025.
  • Level 4 changes the operating model from reacting to losses to tuning controls against a stated risk tolerance, with models, shared intelligence and closed feedback loops.
  • Measure outcomes, not tools: fraud loss rate, false positive rate, customer friction and time to detect matter more than the number of vendors deployed.

What the model covers and where it comes from

The model covers external fraud against the organisation and its customers (account takeover, application and identity fraud, payment and scam fraud, promotion abuse, first-party fraud) and the internal controls that prevent people acting for the organisation committing fraud. It is deliberately vendor-neutral and sector-neutral: a bank, an online retailer and a public body can all use it, with different priority fraud types.

Three public sources shaped the capability areas, and each is used with attribution:

  • GAO Framework for Managing Fraud Risks (GAO-15-593SP): four components, Commit, Assess, Design and Implement, and Evaluate and Adapt. It calls for a dedicated antifraud entity, regular fraud risk assessments that produce a fraud risk profile with a stated risk tolerance, an antifraud strategy focused on preventive controls, data analytics, and outcome-based monitoring.
  • UK Home Office guidance on the failure to prevent fraud offence (Economic Crime and Corporate Transparency Act 2023): six principles of reasonable procedures, namely top level commitment, risk assessment, proportionate risk-based prevention procedures, due diligence, communication including training, and monitoring and review. The guidance says risk assessments are typically reviewed annually or every two years.
  • COSO and ACFE Fraud Risk Management Guide, Second Edition: supports COSO Internal Control Integrated Framework Principle 8 on considering the potential for fraud when assessing risks.

Public bodies in the UK should also read Government Functional Standard GovS 013 Counter Fraud, which the Cabinet Office describes as the minimum public bodies are expected to have in place.

The five levels

LevelNameWhat it looks like
1Ad hocNo named owner. Fraud is handled by customer service, payments or IT as it happens. Losses are written off without root cause. Controls are vendor defaults. Nobody can list the organisation's top fraud types with figures.
2DevelopingA fraud lead or small team exists. Some rules and a verification or device tool are in place for the most visible fraud type. Loss data is collected but categories are inconsistent. Reviews are manual and queues build up during attacks.
3DefinedA written fraud risk assessment and strategy, approved by senior management, covers each priority fraud type with an owner, controls and a risk tolerance. A consistent fraud taxonomy feeds monthly reporting of loss, friction and false positives. Investigation and customer recovery processes are documented. Staff and associated persons receive role-based fraud training.
4ManagedControls are tuned against the stated risk tolerance using confirmed outcome data. Scoring models and rules are versioned, tested in shadow mode and monitored for drift. The team shares and receives intelligence through a consortium or industry scheme. Fraud and cyber teams share signals and incidents. The board receives a fraud risk profile at least twice a year.
5OptimisedFraud prevention is designed into products before launch. Decisions are near real time across channels with adaptive friction. Feedback from chargebacks, complaints, reimbursement claims and investigations retrains models and rules automatically under governance. The organisation contributes to sector typologies and is benchmarked against peers.

Capability areas and level indicators

Use this table to score each area. An area is at a level only when every indicator for that level is true.

Capability areaLevel 2 DevelopingLevel 3 DefinedLevel 4 Managed
1. Governance and ownershipNamed fraud leadExecutive owner, approved strategy, fraud risk appetite statementBoard reporting on fraud risk profile; dedicated antifraud function with authority across product teams
2. Fraud risk assessmentInformal list of fraud typesWritten assessment with likelihood, impact and existing controls per fraud type, reviewed at least every two yearsAssessment refreshed on triggers (new product, new channel, attack wave) and linked to control testing
3. Identity and onboarding controlsBasic document or data checksRisk-based verification by product, synthetic identity and mule checks, device signalsLayered verification with deepfake-resistant liveness, consortium checks, and measured pass and fraud rates per journey
4. Authentication and account protectionPasswords plus SMS one-time codesPhishing-resistant options offered, step-up on risky events, account change alertsRisk-based authentication tuned on outcomes; recovery flows hardened and monitored
5. Transaction and event monitoringStatic rules on some paymentsRules and scores on all priority journeys, scam warnings in payment flows, documented thresholdsReal-time models with shadow testing, drift monitoring and champion and challenger comparison
6. Data, analytics and modelsLoss spreadsheetConsistent fraud labels and a single fraud data storeFeature store, graph analytics for networks, model governance and documented explainability
7. Operations and investigationShared mailbox, ad hoc reviewsCase management, review SLAs, investigation playbooks, evidence retention rulesQueue prioritisation by value at risk, quality assurance sampling, surge capacity plan
8. Customer protection and recoveryComplaint-led refundsDocumented victim support, reimbursement process that meets regulatory rules, account restoration processVulnerable customer indicators, recovery of funds through receiving firms, outcomes reported
9. Intelligence sharing and partnershipOccasional law enforcement reportsMembership of a consortium or industry scheme with a documented legal basisTwo-way sharing with measured precision, platform and telco partnerships, typology contribution

For areas 3, 4 and 9 see the Identity Fraud Prevention Guide, the Account Takeover Prevention Guide and the Fraud Data Sharing and Consortium Intelligence Guide.

Worked example: a mid-size online retailer

A retailer with 400 staff scores itself. Governance is Level 3: a head of payments owns fraud and the board saw a report last quarter. Transaction monitoring is Level 3: the payment provider's scoring runs on every order with documented thresholds. Identity and onboarding is Level 2: sign-up has only an email check, and fake accounts are farming the welcome discount. Customer recovery is Level 1: refunds depend on who answers the complaint. The overall score is not the average of 2.25. Because promotion abuse and refund fraud are its top two losses, the retailer treats itself as Level 1 to 2 where it matters and puts sign-up controls and a documented refund and recovery process at the top of the roadmap, not a new payment model. With more than 250 employees and turnover above £36 million it is also a large organisation for the UK offence, so the plan adds evidence for the six principles.

Self-assessment questions

Answer each question yes or no. A no at a level stops scoring for that area.

  • Governance: Can you name the executive accountable for fraud loss, and when did the board last see a fraud risk report?
  • Risk assessment: Do you have a written fraud risk assessment that is less than two years old and lists a risk tolerance per fraud type?
  • Coverage: For each of your top five fraud types, can you point to the preventive control, the detective control and the owner?
  • Data: Do confirmed fraud cases carry a consistent label that feeds both reporting and model training?
  • Friction: Do you know the false positive rate and customer drop-off caused by each control?
  • Operations: What is your median time from alert to decision, and does it hold during an attack wave?
  • Recovery: Can you show that reimbursement and complaint outcomes meet the rules that apply to you, such as the UK APP fraud reimbursement requirement for payment firms?
  • Sharing: Do you both contribute to and receive from at least one external intelligence source, with a documented legal basis?
  • Internal fraud: If you are a large organisation in the UK, could you evidence each of the six principles of reasonable procedures for fraud committed by employees, agents or other associated persons?

Moving up: a roadmap by level

From Level 1 to Level 2 (first 90 days)

  1. Appoint a fraud lead and agree a simple fraud taxonomy (five to ten types).
  2. Collect three months of loss and case data against that taxonomy.
  3. Deploy or switch on basic controls for the single largest fraud type, for example velocity rules and device signals at sign-up or checkout.
  4. Set up a case log and a shared review rota so alerts are not lost.

From Level 2 to Level 3 (6 to 12 months)

  1. Run a fraud risk assessment workshop per product and channel; record likelihood, impact, controls and risk tolerance.
  2. Write and get approval for a fraud strategy that links each priority risk to controls and owners.
  3. Introduce case management with review SLAs and investigation playbooks.
  4. Start monthly reporting of fraud loss rate, false positive rate, review times and customer friction.
  5. Join one consortium or industry sharing scheme after legal and data protection review.
  6. Deliver role-based fraud awareness training, including for staff and agents whose conduct could expose the organisation to the failure to prevent fraud offence.

From Level 3 to Level 4 (12 to 24 months)

  1. Introduce model governance: version control, shadow testing, drift alerts and rollback for every rule set and model.
  2. Build a single fraud data store and graph analysis for mule and network detection.
  3. Connect fraud and cyber operations so account takeover, credential stuffing and bot signals feed fraud decisions. See the Fraud and Cyber Convergence Guide.
  4. Tune controls against risk tolerance quarterly, using confirmed outcomes rather than alert volumes.
  5. Report a fraud risk profile to the board at least twice a year.

From Level 4 to Level 5

Add fraud review to the product design process, automate feedback from chargebacks, reimbursement claims and complaints into model retraining under governance, and contribute typologies to sector bodies. Few organisations need Level 5 across all areas; target it only where fraud is a core business risk, such as payments, lending and marketplaces.

Small versus large organisations

AspectSmall business or start-upLarge or regulated organisation
Sensible targetLevel 2 in most areas, Level 3 for the top fraud typeLevel 3 everywhere, Level 4 for priority journeys
OwnershipFounder or operations lead with part-time fraud dutyDedicated fraud function with executive owner
ControlsPlatform and payment provider built-in fraud tools, configured rather than defaultedOrchestration of several data sources and models with in-house tuning
Legal driversCard scheme rules, payment provider terms, consumer lawSector regulation, reimbursement regimes, failure to prevent fraud offence if large (two of: more than 250 employees, more than £36 million turnover, more than £18 million total assets)
EvidenceSimple loss log and control listDocumented risk assessment, strategy, testing results and board minutes

Detection and measurement

Maturity claims should be backed by numbers. Track these at Level 3 and above, by fraud type and channel:

  • Fraud loss rate: confirmed fraud loss as basis points of transaction value or as a share of revenue.
  • False positive rate: share of blocked or held events later confirmed as genuine.
  • Friction cost: drop-off at verification and step-up points, and genuine customers declined.
  • Time to detect: median time from first fraudulent event to detection, for each fraud type.
  • Review efficiency: median alert to decision time and backlog during attack waves.
  • Recovery rate: share of victim losses recovered or reimbursed within target timescales.

Framework mapping

Capability areaGAO Fraud Risk FrameworkUK failure to prevent fraud principlesNIST CSF 2.0
Governance and ownershipCommitTop level commitmentGV.RR-01, GV.RM-01
Fraud risk assessmentAssessRisk assessmentID.RA-03
Identity, authentication, monitoringDesign and ImplementProportionate risk-based prevention proceduresDE.CM-09, DE.AE-02
Third parties and partnersDesign and ImplementDue diligenceGV.SC-07
Training and awarenessDesign and ImplementCommunication (including training)PR.AT-01
Measurement and improvementEvaluate and AdaptMonitoring and reviewID.IM-01
Legal and regulatory obligationsCommitAll sixGV.OC-03

Common failure patterns

  • Scoring by tool ownership: claiming Level 4 because a machine learning fraud platform is licensed, while labels are inconsistent and nobody tunes it.
  • Averaging the score: a strong payments team hides an Ad hoc account recovery process, which is exactly where attackers go next.
  • Loss-only reporting: reporting fraud loss without false positives or friction, so controls tighten until genuine customers leave.
  • Stale risk assessment: a fraud risk assessment written for a launch and never refreshed after new products, channels or attack waves; the Home Office guidance warns a court may find an assessment not reviewed recently enough is not fit for purpose.
  • Ignoring internal fraud: treating fraud as only a customer problem and having no evidence of procedures for fraud by employees or agents.
  • No feedback loop: chargebacks, complaints and reimbursement claims never reach the team that sets the rules.

Practitioner checklist

  • Agree the fraud taxonomy and the top five fraud types by loss and harm.
  • Score each of the nine capability areas against the level indicators, with evidence.
  • Set a target level per area for 12 months and get executive sign-off.
  • Write or refresh the fraud risk assessment with risk tolerance per fraud type.
  • Map controls to each priority fraud type, with owners and test dates.
  • Report loss, false positive, friction and time to detect metrics monthly.
  • Check whether the failure to prevent fraud offence applies and document the six principles if it does.
  • Join or review your intelligence sharing arrangements.
  • Re-score every six to twelve months and after any major fraud incident.

Standards and references

Related NHI Mgmt Group resources: Fraud Prevention Guide for Digital Businesses · Fraud Data Sharing and Consortium Intelligence Guide · Identity Security Maturity Model · Fake Account Creation and Sign-up Fraud Guide

PDF for subscribers

Take this guide with you

Every section, table and checklist in one PDF, with links to the sources and related guides. Subscribe to our weekly NHI & AI Identity Journal and we’ll email you a download link.

Free weekly newsletter

Subscribe to the NHI & AI Identity Journal

The latest on NHI and Agentic AI security – articles, research, breaches, news and events every week.

Bonus 33% off our NHI Course when you subscribe.

NHIMG Editorial Note
Researched and written by the NHI Mgmt Group editorial team using AI tools, under the editorial direction of Lalit Choda. Facts are drawn from the sources listed on this page and checked by our automated quality process. Last updated 27 October 2026. Spotted an error? Use our contact page.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM and Agentic AI security. nhimg.org

PDF Downloads for Newsletter Subscribers

Get the guide

Enter your details and we’ll email you a download link.

The full guide is free to read online. The PDF is a thank you for subscribing.

Fill in all fields marked * and tick the PDF box to continue.