How do you know if you have deployed the right automated
screening tools and integrated them successfully? Some indicators you can look for would include
a reduction in the percentage of orders flagged for manual review with no
increase in fraud losses (or even improvement).
A modest decrease in automated rejection rates might also signal
improvements due to reduced false positives (rejection of valid orders).
Another metric to consider is the percentage of orders in
manual review that are approved. From
our experience with merchants, we have seen a high proportion of orders flagged
by automated screening systems but later approved in manual review. Approval rates of 70% or higher are not
unusual. This indicates that the automated
system may not be optimized since many good orders are being forced through
manual review. If this is the case, attempt to correlate automated screening
conditions to good and bad orders, and build rules to automatically handle
conditions based on a high degree of predicted “badness” (fraud prediction)
with a relatively low degree of predicted “goodness” (false positive-ness). Since 50% or more of fraud management budgets
are spent on manual review teams it is critical that this process is
streamlined and efficient.
To get a copy of the 11th Annual Fraud Report, download after registration.
Comments