19 Nov 2026 By CISO Consulting team 15 views 5 min read

SAMA AML Transaction Monitoring Scenario Calibration

Vendor-supplied monitoring scenarios ship with global thresholds that rarely fit Saudi customer behavior. This article sets out three common calibration errors and a back-testing approach based on twelve months of your own transactions.

A financial crime analyst reviewing transaction alert dashboards on two monitors in a modern office.
In this article
  1. What the regulatory expectation actually is
  2. Error 1: Ignoring remittance corridors
  3. Error 2: Applying retail thresholds to SME accounts
  4. Error 3: Calendar-month lookback instead of rolling windows
  5. The back-testing protocol
  6. On sample size
  7. Keep calibration alive
  8. How we support this
  9. Next step

SAMA AML transaction monitoring scenario calibration is where many monitoring programs quietly succeed or fail. A vendor installs a library of scenarios, the thresholds reflect a global customer population, and within weeks the alert queue is full of activity that any analyst familiar with the institution would call ordinary. The team then spends its capacity closing alerts rather than investigating risk.

This article is for Heads of AML and Financial Crime preparing to deploy or tune monitoring rules at a Saudi bank or fintech. It describes three calibration errors that recur in practice, and a back-testing protocol that ties thresholds to your own customer behavior. We do not quote a single figure for how much noise each error causes, because that depends on your portfolio. The method below lets you measure it for your institution.

What the regulatory expectation actually is

SAMA's AML/CTF rules, aligned with the FATF Recommendations, expect monitoring to be risk-based. That means scenarios and thresholds should follow from the institution's own risk assessment, and the institution should be able to explain why a threshold sits where it does. A default value supplied by a vendor is not an explanation. Examiners and internal audit will ask how thresholds were set, what data supported them and who approved them.

Error 1: Ignoring remittance corridors

Saudi Arabia has large expatriate populations sending regular transfers to a limited set of countries. Global scenarios often treat cross-border transfers to certain destinations as uniformly elevated risk, or ignore corridor context entirely. Both approaches produce noise.

Typical symptoms:

  • Repeated alerts on salary-cycle transfers that follow a predictable monthly pattern.
  • Country-risk rules firing on every customer who sends to a common corridor, regardless of amount or history.
  • No distinction between a customer's habitual corridor and a new, unexplained one.

The fix is to calibrate by corridor. Establish what normal looks like for each major corridor: typical amounts, frequency, timing and customer segment. Then alert on deviation from that corridor baseline, while keeping corridor risk, as set in your risk assessment, as a separate factor.

Error 2: Applying retail thresholds to SME accounts

Many deployments use one set of thresholds across personal and business accounts, or a single "customer" segment with a retail-sized profile. An SME account that settles supplier invoices, payroll and customer receipts will routinely exceed limits designed for an individual. The result is persistent alerts on accounts whose behavior is stable and explainable.

The remedy is segmentation before thresholds. At a minimum, separate:

  • Individual retail customers.
  • SME and commercial accounts, ideally split by business type or size.
  • Any segment your risk assessment treats as higher risk.

Thresholds are then set within each segment, so an alert means "unusual for this kind of customer" rather than "large in absolute terms".

Error 3: Calendar-month lookback instead of rolling windows

A scenario that totals activity from the first to the last day of the month has a structural weakness. Activity split across a month boundary is invisible to it, while ordinary month-end behavior such as salary receipts, rent and settlements concentrates into spikes that trigger alerts. Calendar windows create both false positives and missed patterns.

Rolling windows, for example the preceding thirty days measured daily, evaluate behavior continuously and are not distorted by where the calendar happens to fall. Window length should follow the typology the scenario targets: short windows for rapid movement of funds, longer ones for gradual structuring. Document the reasoning for each choice.

The back-testing protocol

Calibration should be evidence-based. The approach we use relies on twelve months of your own transaction history, long enough to include seasonal peaks such as Ramadan, Hajj and Umrah periods and year-end activity.

  1. Define the baseline. Draw twelve months of transactions and attach customer segment, corridor and risk rating from your risk assessment.
  2. Profile behavior by segment and corridor. Calculate the distribution of amounts, frequencies and velocities for each combination that matters.
  3. Run candidate thresholds against history. Apply each scenario with rolling windows and see how many alerts it would have produced at several threshold levels.
  4. Review a sample of the alerts. Have experienced analysts assess whether the alerts at each level were genuinely unusual. Use the outcome to judge whether the threshold separates deviation from normal behavior.
  5. Test known cases. Confirm that previously identified suspicious activity, and relevant typologies, would still have been detected.
  6. Set, approve and document. Record the chosen thresholds, the evidence, the residual risk and the approver.

On sample size

There is no single number that is correct for every institution. The sample must be large enough that each segment-and-corridor combination you calibrate has enough observations for its distribution to be meaningful, and the alert review sample must be large enough per scenario for analysts to judge precision with confidence. Small segments may need to be grouped or handled with more conservative thresholds and closer review. Whatever you choose, record the sample size and your reasoning, because that record is what demonstrates that thresholds match actual customer behavior.

Keep calibration alive

Thresholds drift as products, customers and corridors change. Set a review cadence, define triggers for out-of-cycle tuning such as a new product launch, and keep a change log with approvals. Model governance is part of the control, not an afterthought.

How we support this

Calibration works best when designed into the implementation rather than repaired afterwards. Our AML Compliance Solution covers vendor-independent selection and implementation of AML technology, including customer risk scoring, sanctions screening, transaction monitoring and case management. The work follows four stages: define requirements, select the vendor, implement data integration, scenarios and risk models, and validate before go-live.

The vendor implements the software; we define, oversee and validate on your behalf, so you have independent assurance. Deliverables include scenario and model design documents and a validation report and governance pack, which give you the evidence trail described above. Projects typically run 4–9 months.

Next step

If you are about to deploy monitoring or tune an existing set of rules, start by checking your scenarios against the three errors above. Then decide whether your thresholds can be defended with your own data. Our AML Compliance Solution is the place to start that conversation.

Tags: SAMA AML/CTF transaction monitoring scenario calibration false positives financial crime back-testing
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