How to optimise transaction screening: Fewer alerts, improved accuracy
Transaction screening can be a source of headaches for fincrime teams at banks and fintechs, with a high proportion of false positives meaning that unnecessary time is wasted on manual reviews. This can strain resources and create unnecessary friction for legitimate customers.
In this scenario, reducing the number of false positive alerts that require manual review is key. By tuning out the noise and focusing on the transactions that really matter, it’s easier to identify more fraudulent transactions and, as a result, beat more criminals.
Ester Eggert, Head of Product at Salv, recently hosted a webinar on how you can reduce the volume of transactions in manual review to less than 0.1% without compromising compliance standards.
The four steps in this blog will help you kickstart the search for that 0.1% Utopia. But if you want to get (even) more insights, then watch the full webinar recording here.
Be laser-focused when choosing the right metrics to measure
The first step in this process is setting the right KPIs. “If you don’t know what you’re measuring, then it would be very difficult to say whether you are successful or not,” says Esther.
One challenge that fincrime teams frequently face is that there are many metrics to measure, which leads to disagreements on what is relevant. Ester says that there’s no perfect set of KPIs that can capture every angle. Selecting a dozen or so KPIs is not a solution either, as it increases the risk that teams will lose focus on what is important.
Ester points out that the maximum number of KPIs teams should track is three. Two KPIs is optimal because this will streamline focus even more, and the two that we recommend tracking are:
- The ratio of transactions that get stuck in manual review: Monitoring ratio is helpful because it highlights your scalability. The lower the ratio, the easier it is to scale transaction volumes.
- How meaningful your alerts are: If something takes two seconds to resolve, it probably wasn’t meaningful and could have been handled without human intervention.
Standardising data can help cut false positives
Data quality is key to successful transaction screening. “In many cases, you will not have the power to influence the quality of the data that comes in,” says Esther, but there are steps that can be taken to improve the consistency and reliability of that data.
This means attaching as much context and additional information as possible to transactions within the screening engine. Where possible, compliance teams should also standardise data to help reduce false positives. Some examples of this are ensuring the format of first and last names is consistent, and avoiding initials and titles, which can often confuse the algorithms, Ester says. The same also applies to company suffixes.
Highly configurable systems can improve screening flexibility
Compliance teams must be capable of fine-tuning and configuring how the screening engine works. By putting the power and flexibility in the hands of the teams using it (and not the IT department) you can apply updates as quickly as required, without waiting for another department.
Having a highly configurable system not only means compliance teams can adjust their risk appetite as circumstances dictate, it can also drive efficiency. If you configure your engine properly, you can reduce your workloads by a lot,” says Esther. For example, having the flexibility to apply different screening rules across different countries can be useful if regulations allow firms to wave through transactions under a certain threshold. This means less work for your team and less friction for customers.
For more insights on optimising your transaction screening, watch the full webinar with Ester here.
Using automation to reduce manual intervention
The final stage of this process is automating decision-making to cut unnecessary manual reviews. Take updates to sanctions lists. “Often there are some minor updates in sanctions lists which abolish good listing,” explains Esther. If you can define what constitutes a significant update, you can ignore anything minor or irrelevant and ensure goodlists are only overridden with important new information.
This also applies to counterparty-level goodlists, which may or may not flag a customer’s transactions based on the type of entity the counterparty is. Ester adds that you should also be looking to auto-resolve obvious false positives by using secondary information and known patterns, while also being able to adjust decision automation based on changing risk appetites.
By following these four steps, you can reduce your manual intervention time and ensure the alerts that get flagged are meaningful and worth investigation. For more detailed insights, watch the full webinar with Ester here.
