If you’re plagued with the problem of high levels of false positives, you’ll know they are a big and costly problem.
And one that needs prompt attention.
With average false positive rates for legacy sanction screening systems as high as 5-8%, we estimate false positives can cost an eye-watering £200,000 per million customers/transactions screened. This is based on the number of transactions and the time taken to manually review each one.
Reducing false positives is therefore absolutely essential if the time, resources and money of compliance departments is to be well spent. Low levels of false positives will enable efforts to be better focused on analysing true hits.
While it’s a very difficult task to completely eliminate false positives when taking a balanced risk-based approach to sanction screening, each of the below are sure-fire ways to see a safe and welcomed reduction:
1. Capture data in a clear, structured way
With so much data to be processed and analysed, data must be captured in a clear, structured way. This means, for example, gathering title, first, second, and last names in different fields. Not only does this avoid any potential ambiguity, but it makes it easier to match against an external sanction or data list.
With careful data capture design, you can minimise the risk of a data capture error. This is important because even something as simple as inserting a first name into a surname field could trigger a false positive.
2. Use as much data as possible
Criminals and terrorists will do their best to beat anti-money laundering systems and remain undetected. Tactics like changes of name and moving from one country to another are common.
For this reason, the more relevant data you have for verification, the bigger and more accurate picture of their identity and history you can create. More importantly, the more relevant data you have for verification, the lower the risk of false positives and false negatives occurring.
3. Improve matching algorithms
In order to reduce false positives and false negatives, take care when combining matching algorithms and configuring them based on language, scenarios and company policies. You should use different matching algorithms that account for different cases. These should be weighted based on the scenario and contribute to the overarching score that determines the sanction risk.
For example, for a company name, a higher importance can often be placed on the first name in comparison to the other names, which are typically more common and less relevant when it comes to matching. Consider, for example, Test123 Management Consulting Limited and Test360 Management Consulting Limited.
The ability to calculate average scoring and standard deviation is also beneficial as it may give important information on whether different approaches “agree” in considering whether an alert is a false positive or a true hit.
4. Continuously improve your screening process
Reviewing the output of your screening process, including hits and false positives, may help improve the rules and scenarios that have been configured in your system, or potentially help you improve the policies you are adopting.
5. Use whitelists to your advantage
The opposite of blacklists, whitelists can be used to support sanction compliance efforts by identifying and saving repetitive matches that materialised as a false positive.
That said, while whitelists can help reduce the occurrence and cost of false positives, they can increase the risk of false negatives. In the case of a sanction breach, it can be difficult to satisfactorily explain to regulators why further investigations weren’t made. Whitelists should therefore be used with caution, and only as a means of supplementing screening processes.
6. Use alternative scoring
Traditional distance-based scoring alone can lead to excessive levels of false positives because thresholds are set manually. Using alternative scoring in addition to distance-based scoring, however, can complement and enhance the screening process, leading to more reliable results.
For example, machine learning generated classification scores can take into account different dimensions within a match, such as average length of words, average similarity score and maximum similarity score.
The above 6 steps are really effective but bear in mind you will only be able to reduce false positives as far as your current sanction screening systems and processes will let you.
To see a substantial difference, you may want to consider upgrading your current system with the very latest sanction screening technology.