Use the Route Metrics reports on the workflow configuration page to understand workflow performance and make adjustments. Below are some common scenarios to watch out for.
Low precision routes that can be restricted
In this example, we’re looking at the Accuracy tab of a review queue route. Note the high percentage of users that resulted in an accept decision (the blue sections). This indicates that the route can be further restricted. At the very least, you could set a score threshold somewhere between 5 and 20 (outlined in gray above). Below 5, all users have been manually accepted. Below 20, no users have been manually blocked.
High-precision routes that can be expanded
If an auto-blocking route results in a very high rate of confirmed fraud (manual block decisions), this is an indication that you can likely lower the blocking threshold and still maintain high precision.
High overturn rates
A decision gets “overturned” when a blocked case is later marked with an accept decision (a false positive), or when an accepted case is later marked with a block decision (a false negative). In this blocking route, 30 out of 60 users who passed through ended up getting overturned - this is a relatively high overturn rate and may indicate that the route needs to be updated to improve precision.
Opportunities to automate more and review less
Note the high ratio of red (manual block decisions) in the 53+ score range. This indicates an opportunity to auto-block using the same route logic and a score threshold of 53, reducing some of the cases that would otherwise get routed to manual review.
Routes that don’t trigger
Take note of any routes that show Runs Started as 0 or a very low number. If a route has been triggered very few times or not at all, consider removing it to clean up your workflow.