Rule tuning
Rule tuning is the ongoing adjustment of fraud rule conditions and thresholds to catch more genuine fraud for the alert volume analysts can handle. It is done by testing a change against past transactions first, so its effect is known before it affects live work.
In more detail
Tuning works on two dials. Loosening or tightening a threshold moves the balance between missed fraud and false positives, and adding or removing conditions changes how narrowly a rule targets its pattern. Each change has a cost, so tuning is a decision to be evidenced, not a hunch.
The evidence comes from replaying a proposed change against historical transactions and comparing what it would have flagged, then checking results afterwards. The misconception is that tuning is a one-off setup task. Fraud patterns shift and customer behaviour changes, so a rule set that is not revisited steadily loses accuracy.
Each candidate change is replayed against stored transactions, which records what the old and new versions would have flagged, and the analyst compares the two before adopting it.
Example
An analyst proposes raising a velocity limit after seeing most alerts cleared. Replayed over last quarter, the change would have removed a large share of the cleared alerts while still flagging every confirmed fraud case, so it is adopted.
Related capability
Rule changes and versioning in the rules engine
Related terms
Discuss your transaction monitoring requirements
Tell us which transaction flows you need to monitor, which systems Akoben would integrate with, and how your fraud team works today.
Request a demo