Screening tuning: what the decision maker looks at

Screening tuning adjusts the thresholds and matching rules behind sanctions, PEP and adverse media screening so that alerts reflect real risk rather than noise. The decision belongs to the compliance officer, not to the screening vendor or the analyst who raised the request. Tuning follows a fixed sequence: baseline review, proposed change, test run, sign-off, and monitored rollout. Skipping a stage is the most common finding in later reviews.

What changes once tuning is requested

Once a tuning request is logged, the alert queue does not change immediately. The existing configuration keeps running until the new parameters are tested and signed off. This matters because an obliged entity remains exposed under the old configuration during the whole review window, and any match generated before rollout is handled under the rules that were active at the time it appeared.

A second, quieter change is documentation weight. Every tuning request creates a file: who asked, why, what evidence supported the request, and who approved it. That file becomes the first thing an auditor or a supervisor asks for if a later alert is missed. Tuning without a file is treated as an undocumented control change, regardless of how sound the underlying reasoning was.

Steps to take

1. Log the request in writing. The analyst or system owner states the observed problem — too many false positives on a specific name pattern, or a missed match discovered after the fact.
2. Pull the baseline metrics. The compliance officer requests alert volume, false positive rate and average review time for the affected rule over a defined period, usually the preceding quarter.
3. Draft the proposed change. The exact threshold, fuzzy-matching parameter, or exclusion list entry is written out, not described in general terms.
4. Run the change in test mode. The new configuration runs in parallel with the live one for a set window, without suppressing any live alerts.
5. Compare outcomes. The compliance officer reviews what the new configuration would have missed and what it would have added, side by side with the old one.
6. Sign off and roll out. Approval is recorded with a name, a date and the comparison data attached, not as a verbal go-ahead.

Who decides and how long it takes

The compliance officer, or a named deputy with documented delegation, makes the final call. A vendor recommendation or an analyst's preference is input, not a decision. In smaller organisations the same person may run the test and sign off; in larger ones a second reviewer checks the comparison data before approval.

The point of no return is rollout, not sign-off. Before rollout, the old configuration can be reinstated without consequence. After rollout, every alert — or absence of one — is measured against the new configuration, and reverting requires the same documented process in reverse.

Common branches

If the tuning request comes from a single missed match rather than a volume complaint, the review widens. The file then has to explain not just the rule change but why the earlier configuration allowed the miss, and whether other alerts from the same period need a second look.

If the organisation holds more than one licence and tuning affects a shared screening engine, the change touches every licensed activity at once. The same escalation logic applies elsewhere in the file: An authorisation is withdrawn authorisation who decides and when follows the same principle — a named person decides, and the record shows why.

If the tuning affects screening tied to a portfolio of registered rights rather than customer names, the ownership question comes first. That is the same question that arises in When an assignment breaks the chain of title: before a rule is changed, someone has to confirm who is entitled to change it.

If a counterparty disputes the outcome of a screening decision and escalates it, the file may end up feeding a formal claim. The decision point there resembles When an injunction application is served: the clock starts on receipt, not on internal review of merits.

If a client or counterparty invokes a contractual audit clause during the tuning window, the review has to pause and produce the file on demand. This mirrors When an audit right is invoked without notice: the request does not wait for the tuning cycle to finish.

What this page does not cover

This page describes the internal decision procedure for tuning a screening configuration. It does not set out which screening tools or thresholds are appropriate for a given size of business, and it does not describe how a suspicious activity report is filed once a genuine match is confirmed. That question is addressed separately: Suspicious activity reporting step by step covers what happens after a match is treated as real rather than a false positive.

Verified

This page describes an internal control procedure rather than a statutory obligation, so no specific legal provision is cited. The sequence of stages, ownership and typical timing reflects standard internal-control practice for obliged entities and is reviewed against current practice at the time of publication.

Frequently asked questions

How long does screening tuning take in practice?

From request to rollout, a typical tuning cycle runs from three to six weeks. The parallel test run usually accounts for most of that time, because the comparison needs enough live data to be meaningful.

Which documents are needed to start?

A written request stating the observed problem, baseline metrics for the affected rule, and the name of the person who will own the sign-off. Without these three, the compliance officer cannot open a comparison window.

What happens if the other side does not respond?

If a customer or counterparty does not respond to a clarification request during a screening review, the alert stays open. It is not closed on the assumption of an eventual answer, and the file records the outstanding request until it is either resolved or escalated internally.

— Aleks Vaher, AURELIS