Identify Optimum Threshold Parameters


Threshold Analyser is a new analytical approach to identifying optimum threshold parameters when tuning a sanction screening system. 

Our new product will model a decrease in the number of false positives produced by the system without reducing accuracy, thus ensuring maximum resource efficiency at specified threshold levels.

An interactive threshold level selection toggle can be moved in Analyser Online to view the impact on efficiency and effectiveness levels with each selected threshold change. 

Applying a risk-based optimisation approach will lead to an accepted balance between maximum sanction alert hit rates whilst ensuring maximum resource efficiency at specified threshold levels.

From the returned information from the tested system, AML Analytics are able to link the test record back to the name that was matched and the threshold score at which it was matched.

We then determine match rates to Control, Manipulated and Clean ID test records at each threshold level to analyse the impact on scores if system matching thresholds are decreased or increased. 

The relationship between the number of returns and alerts missed per threshold value is used to determine the optimum threshold value. It can be seen that the number of returns will decrease as the threshold is increased, and then conversely, the number of alerts missed will increase as the threshold is increased.

From these metrics, calculate your optimum threshold level whilst taking into account specific risk appetite and aim to improve efficiency scores without adding any additional sanctions risk. 

To find out more, contact a member of the AML Analytics team by clicking on the interactive map

Optimality Calculator


The Optimality Calculator can be used to model potential savings on resource in terms of both time and cost for a financial institution at various specified threshold levels.

If the number of low-quality alerts is reduced then fewer analysts will be needed to check these alerts and a high standard of alert review will be maintained.

Model exactly what these cost savings will be for your organisation at any threshold level you wish by entering your own costs into the Optimality Calculator and setting a suggested threshold level to view results immediately.

To find out more, contact a member of the AML Analytics team by clicking on the interactive map

Control and Manipulated Explained

Control tests are created by using records exactly as they appear on sanction lists and without making any changes to these names. We refer to these datasets as Control tests which are designed to test the pure matching capability of a system. A Control test may include unchanged primary names and aliases of individuals, entities vessels, aircraft, countries, territories and BICs (Business Identifier Codes) on sanction lists.

Manipulated tests include records that have been changed using algorithms from the way they appear on sanction lists in order to copy how a criminal might change their name or to emulate mistakes made at data entry level. Using Manipulated data in tests will test the fuzzy logic matching capabilities of a sanction screening system.  It is normally harder for a sanction screening system to hit against records in a Manipulated test than in a Control test.