20 Nov 2026 By CISO Consulting team 10 views 4 min read

Fraud Management System RFP: 20 Requirements to Test in a PoC

Vendor demos show polished dashboards, but only a proof of concept on your own data shows whether a system fits. This article sets out twenty requirements to test so you can score vendors on equal terms under SAMA CFF expectations.

A fraud analyst reviewing transaction alerts on a monitor in a modern operations room.
In this article
  1. How to set up the PoC
  2. Detection and latency
  3. Rules and models
  4. Case management and evidence
  5. Integration, security and operations
  6. Turning the list into a scoring sheet
  7. How we can help

A fraud management system selection RFP often rewards the vendor with the best presentation rather than the best fit. Demonstrations run on clean sample data, prepared scenarios and a polished dashboard. A proof of concept (PoC) on your own data, with your own analysts, shows what the system will do on a normal Tuesday and on the worst day of the year.

This article sets out twenty requirements worth testing in a PoC, grouped by theme. Use them to write a scoring sheet that compares vendors on equal terms, with evidence for every score. The context is the SAMA Counter-Fraud Framework (SAMA CFF), which expects banks and payment firms to detect, investigate and report fraud in a governed and demonstrable way.

How to set up the PoC

Before testing anything, fix the conditions so that every vendor faces the same test.

  • Give each vendor the same data extract, covering the same channels and period, with sensitive fields masked as your data-protection rules require.
  • Include known fraud cases that your team has already confirmed, so detection can be measured against a truth you trust.
  • Agree the scoring sheet before the PoC starts, with weights that reflect your risk profile.
  • Have your own analysts operate the system, not only the vendor's engineers.

Detection and latency

  1. Real-time latency. Measure end-to-end decision time at your expected peak volume, not at average load. Record the slowest responses as well as the typical ones.
  2. Channel coverage. Test cards, mada, SARIE, instant payments and digital channels with realistic message formats. A gap in one channel is a gap in the control.
  3. Detection on known cases. Replay confirmed fraud and see which cases the system flags, how early, and with what score.
  4. False-positive burden. Count the alerts your analysts must clear for each confirmed case. A system that catches fraud by drowning the team is not a fit.
  5. Behavioural and profile data. Check whether customer and device history is used in the decision, and how quickly profiles update after a new event.
  6. Behaviour under degradation. Simulate a slow or unavailable data source and observe whether the system fails open, fails closed or degrades by a rule you control.

Rules and models

  1. Rule authoring without vendor tickets. A trained fraud analyst should be able to write, test and deploy a rule within your change process. Time how long it takes, and note whether any step needs the vendor.
  2. Simulation before deployment. Rules should be back-tested against historical data to show expected alert volume before they go live.
  3. Model explainability. For every score, the system should show which factors drove it, in terms an investigator and an auditor can follow.
  4. Model governance. Look for versioning, approval, performance monitoring and a way to retire or roll back a model.
  5. Champion and challenger testing. The system should let you run a new rule or model alongside the live one without affecting customers.
  6. Rule portability and ownership. Establish who owns the rules and models you build, and in what format you can export them.

Case management and evidence

  1. Alert-to-case workflow. Test assignment, prioritisation, escalation, and service-level tracking with your own team structure.
  2. Investigator context. An analyst should see the transaction, customer history, linked alerts and prior outcomes on one screen without switching tools.
  3. Alert-to-evidence trail. For a sample of alerts, trace the path from trigger to decision, including who acted, when, and on what information. This is the evidence a SAMA CFF review will ask for.
  4. Outcome feedback. Confirmed fraud and false positives should flow back into tuning. Check that the loop exists and is not a manual export.
  5. Reporting and regulatory output. Produce the management and regulatory reports you need from the system itself, with data you can reconcile to source.

Integration, security and operations

  1. Integration effort. Note what the vendor needed from your teams to connect the data sources, and how many open questions remained at the end. This is the best predictor of implementation effort.
  2. Security and access control. Check role-based access, segregation of duties, logging of administrator actions and handling of card data in line with PCI DSS where it applies. Confirm data residency and hosting terms against your regulatory obligations.
  3. Operational resilience and support. Review availability design, failover, upgrade process and support response, and ask for these in contract terms rather than in a presentation.

Turning the list into a scoring sheet

Give each requirement a weight, a defined test and a scoring scale from 0 to 4 with a written description for each level. Keep three things separate: mandatory requirements that disqualify a vendor if missed, weighted requirements that drive the ranking, and commercial factors that are scored independently. Record evidence for each score, such as a screenshot, a timing log or a replay result, so the decision can be defended to a committee, to internal audit and to SAMA.

Common gaps we see in selection exercises are weights set after the demonstrations, PoC data that is too clean, and rule authoring tested by the vendor instead of by your analysts. Each of these favours the vendor over the buyer.

How we can help

Our Fraud Management Solution service is a vendor-independent selection and implementation of a fraud management system, covering requirements, vendor evaluation, rules and models, integration and go-live. It runs in four stages: define, select, implement and go live. The select stage covers the RFP, demonstrations, proof of concept and scoring.

The deliverables include a requirements specification, a vendor evaluation report and recommendation, implementation oversight reports and go-live test evidence. The project typically runs 4–9 months, and it is built around SAMA CFF and PCI DSS.

We do not resell any fraud system, and we declare any relationship before an evaluation starts. If you already have a system, we can start at implementation or optimization rather than selection. To understand your current exposure first, a fraud assessment scored against the SAMA Counter-Fraud Framework is a sound starting point.

Tags: fraud management system RFP proof of concept SAMA CFF financial crime vendor selection
Readiness self-check

How ready are you? Find out in two minutes

Answer a few questions for your framework. You get a score, your biggest gaps and — if you want it — a detailed assessment from our team.

Ready to talk about your compliance?

Tell us where you stand. We will show you the shortest path to what your regulator expects.

Regulatory updates in your inbox

SAMA, NCA and SDAIA changes and what they mean for your institution — once a month.

E-mails in

We confirm by e-mail; unsubscribe any time. Manage your subscription · Past issues

Schedule a Free Assessment