13 Nov 2026 By CISO Consulting team 26 views 5 min read

Sanctions Screening Software Selection Checklist: 12 Names

Vendor demos use clean sample data, so they reveal little about how a tool handles your real customers. This article shows how to build a twelve-name test file and score each vendor on what it finds, misses and over-alerts.

Compliance analysts in a Saudi bank office reviewing screening results on a large monitor during a vendor evaluation.
In this article
  1. Why demo data tells you so little
  2. Signs it is time to run a structured test
  3. Building the twelve-name test file
  4. Group 1: Arabic transliteration variants (four names)
  5. Group 2: Name-order and structure differences (three names)
  6. Group 3: Partial matches (three names)
  7. Group 4: Entities and common names (two names)
  8. Setting the expected result before the demo
  9. Scoring each vendor
  10. From demo to proof of concept
  11. Framing the result for the committee
  12. How we run this
  13. Next step

A sanctions screening software selection checklist is only as useful as the evidence behind it. Vendor demos run on clean sample data, where names are spelled one way, in one order, in one script. Your customer base is not like that. This article describes a small test file of twelve names that you can run in every demo, so that each vendor is judged on the same evidence and the selection committee receives a result it can defend.

Why demo data tells you so little

A demonstration is designed to succeed. The names are chosen because the engine handles them well, and the alerts shown are tidy. What the demo rarely shows is how the tool behaves with Arabic names written in Latin letters in several accepted ways, with names whose order differs between a passport, a core banking record and a sanctions list, or with very common names that generate noise.

For heads of compliance and MLROs at banks, finance companies and exchange houses, the practical questions are simple. Does the engine find the person it should find? Does it miss anyone? How many alerts will analysts have to close that mean nothing? A test file answers all three with your own data rather than the vendor's.

Signs it is time to run a structured test

  • You are replacing an engine that produces alert volumes your team cannot clear.
  • Analysts report that known variants of a name are not being matched.
  • Your sanctions program now covers UN, local, OFAC and EU lists, and the current tool handles some of them unevenly.
  • Procurement needs a scored comparison, not impressions from presentations.

Building the twelve-name test file

Keep the file small enough to run live in a demo and to review by hand afterwards. Use names that are already public on the lists you screen against, or fictional names built to mimic them. Do not load real customer records into a vendor environment without your data-protection approval.

Group 1: Arabic transliteration variants (four names)

Choose listed names that have several common Latin spellings, for example differences in how the definite article, the letters representing ayn or qaf, and prefixes such as "Abd" are written. Include at least one name with a hyphenated or split compound form. Submit the variants a customer record might plausibly contain, not only the spelling on the list.

Group 2: Name-order and structure differences (three names)

Test a listed name with the family name first, the same name with the given name first, and one with a middle or patronymic element missing. Include a name with a title or honorific attached. A sound engine should treat these as the same person without being told.

Group 3: Partial matches (three names)

Submit a name with one element omitted, one with an initial in place of a full element, and one with a minor typographical error. These show how the tool balances sensitivity against noise, and whether the threshold can be tuned.

Group 4: Entities and common names (two names)

Include one listed entity whose trading name differs from its legal name, and one very common personal name that sits close to a listed individual. The common name is your false-alert test. It shows what happens when many harmless customers resemble one listed person.

Setting the expected result before the demo

Write down, before any vendor sees the file, which names should produce an alert and which should not. Record the list entry each should match. This answer key is what turns the demo from a performance into a measurement. Without it, every result can be explained away afterwards.

Scoring each vendor

Use the same sheet for every vendor and complete it during or immediately after the session.

  1. Found: how many of the names that should alert did alert, and against the correct list entry.
  2. Missed: which expected matches did not alert, and whether the vendor could explain why.
  3. False alerts: how many alerts fired on names that should not match, and how clearly the tool showed the reason for each match.
  4. Tuning: whether thresholds, fuzzy logic and name-handling rules can be adjusted by your team without vendor involvement.
  5. Evidence: whether results and match rationale can be exported for audit and for the committee file.

Weight misses more heavily than false alerts. A missed true match is a regulatory exposure; a false alert is an operating cost. Agree the weighting with the committee before scoring begins, so it cannot be adjusted to favor a preferred vendor.

From demo to proof of concept

Twelve names is a screening filter, not a final test. Vendors that perform well should move to a proof of concept using a larger, representative sample of your own data, handled under your data-protection controls. Repeat the same answer key logic at greater volume, and add the list-update process, the handling of the UN, local, OFAC and EU lists, and how the tool fits with case management.

Framing the result for the committee

The committee needs a short, comparable record: the test file, the answer key, the weighting, each vendor's scores, and a plain statement of the residual risks. Where a regulator such as SAMA reviews your AML/CTF controls, this record shows that the choice was reasoned and evidence-based, consistent with the FATF Recommendations on targeted financial sanctions.

How we run this

Our AML Compliance Solution is a vendor-independent selection and implementation service covering customer risk scoring, sanctions screening, transaction monitoring and case management. The work follows four stages: define, select, implement and validate.

In the define stage, we set requirements across screening, monitoring and case management. In the select stage, we run an RFP, demonstrations, a proof of concept and scoring. That is where a test file like this one is used, as part of a fair, scored vendor evaluation. The outputs are a requirements specification and a vendor evaluation report with a recommendation.

Implementation covers data integration, scenarios and risk models, and the final stage is pre-go-live validation with a governance pack. The vendor implements the software; we define, oversee and validate on your behalf, so you have independent assurance. The project typically runs 4–9 months, and it can include replacing a current system, with migration of history, rules and open cases.

Next step

If you are preparing a shortlist, start by drafting your twelve names and the answer key. If you would like support building the file, running the demonstrations and scoring the results, our AML Compliance Solution is designed for that work. Contact us to discuss your requirements.

Tags: sanctions screening AML technology vendor selection Arabic transliteration SAMA proof of concept
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