Are you suffering from false positivity?
We’re talking endless barrages of faux-important alerts that clutter your team’s work queues and drag down productivity. Spending time to review an alert just to realize it’s not valuable is a pain point all too familiar to Alert Investigators. This can significantly reduce the costs to your AML compliance operations, saving you time on investigating what really matters.
That’s why Unit21 created Alert Scores - to help you focus investigator time on the alerts with the strongest signal.
What are Alert Scores?
Alert Scores work on a scale of 0-100 to provide a numerical value that fraud agents can easily interpret. Prioritize the most high-risk alerts to be effective.
Alert Scores are based on a vetted machine learning module that is trained on your prior alert dispositions and behaviors. Using the Unit21 platform will passively train your organization’s machine learning model, without any extra necessary actions.
How Does This Benefit My Team?
Alert Queues can be easily sorted by alert scores, so only the highest priority alerts get the attention they deserve. This makes the process of working through alerts easier and reduces false-positive rates to free up your team’s valuable time.
For example, for some teams, the lowest-scored alerts can be ignored as false positives with an accuracy of >99%, with the remaining alerts having a true positive rate of >40%.
With more time, you can set more ambitious rules, and as you gain more confidence in investigations. Also, reduce “Average Handle Time” for alerts with Unit21 provided signals.
Boost productivity while onboarding new agents. With a simple, user-friendly UX that clearly highlights which alerts need attention, inexperienced agents can be more self-sufficient in investigating alerts of appropriate complexity.
Alternately, select lower-scoring alerts for onboarding or junior agents while reserving high priority investigations for seasoned fraud or compliance officers. Finally, drive more operational efficiency with escalated alerts to senior investigators for faster SAR filing.
Moreover, unlock interesting workflows by working alerts you usually wouldn’t. Instead of having traditionally “best-effort” or normally backlogged queues, use alert scores to help you disposition high-priority alerts faster.
Then, use the time saved to look over experimental or control (random or intelligent sampling-based) rules sent to different queues as you continue to refine your rules and provide additional training signals to Unit21’s Alert Score models. Reviewing more behavior, in general, allows more confidence and accuracy of Unit21’s model.
How to Start using Alert Scores
Make your agents happier with immediate numerical (and visual - the UX will become red when an alert score is high, signaling a high priority alert) indicators to agents when working on an alert with strong signals based on a machine learning model trained prior alert dispositions.
If you would like access to these scores in your dashboard, get in touch, and we will set you up!
By: Andrew Ardito
Andrew Ardito is a Sr. Product Manager who works on Unit21’s machine learning initiatives.