Evaluate New Fraud Tools

Optimization Engine helps Credit Card Issuer determine ROI of new fraud tools before investment.

Fraud Tags

data elements that identify if an application was labeled as fraudulent or not

Indicators

elements of a fraud targeting strategy used to identify fraudulent transactions. Indicators can be alerts, business rules or scores

Individual Performance

how an indicator would perform if it was the only indicator

Incremental Performance

the incremental benefit an indicator provides after accounting for all other indicators

Retrospective Analysis

using historical applications with known outcomes to simulate “what if” scenarios

Background

A credit card issuer uses a combination of internal fraud scores, a third party fraud score, 25 business rules and a 3rd party identity verification alert to rank every loan application they receive by fraud risk. They are considering purchasing a new email-based identity verification tool to add to their fraud detection strategy. Before they invest in this new tool they want to run a retrospective analysis to determine how much fraud this tool would have detected and what incremental fraud it would have helped prevent, if any.

The Process

The new vendor being evaluated agrees to support a free retro analysis on 6 months of historical applications. Outlier Analytics helps the credit card company prepare an application data file to send to the vendor. The file does not contain fraud tags or performance information. The vendor appends their alerts and risk scores and returns the file. The credit card company then appends the fraud tags and information on which rules were triggered for each application. The file is securely transferred to Outlier Analytics where it automatically is processed through the Optimization Engine. Within seconds performance results are provided showing fraud detection and false positive rates for every risk score segment, business rule and alert. The results provide individual performance metrics and rankings based on incremental performance.

Results

Using results from the Optimization Engine the credit card company draws the following conclusions on the new vendor solution:

shows great individual performance, a top 5 indicator detecting 14% of the fraudulent applications
provides little incremental benefit beyond the current indicators being used, 2% incremental fraud detection with the new solution compared to without it
the incremental fraud detection comes with a high false-positive rate which would require additional analysts to cover the increased review volume

Based on the results observed from the retro analysis the credit card company decided not to invest in the new vendor solution.

Optimization Benefits

Simulate unlimited scenarios with the Optimization Engine by adding, removing, and/or combining indicators and use results to quantify impacts and calculate ROI of strategy changes.