
Focusing on good customers is key to a more effective AML transaction monitoring process
Transaction monitoring has become an endless arms race turning out very costly, cumbersome and which produces very little concrete results.
It’s time for a rethink. Instead of fighting a losing battle trying to model criminal behaviour, as we’re used to, why don’t we model the behaviour of good customers?
Training a model to catch financial criminals is like trying to find a needle in a haystack – except the needle is constantly changing to avoid being found. A much more effective approach is to train your model to recognise the hay.
After all, there’s substantially more hay than needles and it has no incentive to change, making it much more predictable. The model therefore doesn’t need to know what needles look like – it just needs to know whether something isn’t hay.
Just like any person finding out about a structural crime saying “why didn’t we see this one, it looks very different from what you would expect in this situation”. Well, they weren’t looking for crime in that way, only focusing on known criminal phenomenon.

This is why we talk about modelling good behaviour. What can you expect in a situation from a totally legitimate customer?
Look at the haystack, not the needle
Instead of trying to spot criminals, this approach focuses on normal customer behaviour with a model that’s trained on legitimate transaction data. This might sound simple, but it works. Looking at customers instead of criminals shines a light on previously undetected patterns. Banks that have implemented this type of model have often substantially increased their number of previously undetected crime and cut their false positives by 80%.
This rethink also brings other benefits. Instead of relying on dozens of incomprehensible rules, being tweaked and tuned over time, this risk-based approach is more straight forward and therefore easier to explain to regulators. Think white box, not black box.
Compliance costs
The stakes are high. Most news on transaction monitoring has been about banks incurring fines, but that’s only the tip of the iceberg. LexisNexis estimates the global cost of AML and sanctions compliance to be more than $210bn a year. Most of this money is spent on cumbersome administrative processes targeting totally legitimate customers while doing their daily business.
Despite these staggering efforts, criminals continue to thrive by finding the loopholes in the system. It simply seems impossible to keep pace with knowing how they operate, turning the efforts in an endless cat-and-mouse game with very limited results.
By focusing your model on the good customers, you can focus your efforts again on the real criminals, rather than the other way around.
Start modelling good behaviour today

Want to know more about how this approach can increase protection and drastically reduce false positives of your existing monitoring system?
Get in touch and we’ll be happy to discuss solutions.