September 12, 2022

When it comes to fighting financial crime, larger financial institutions can have a competitive edge with slick and expensive enterprise fraud solutions. When increased end-user friction leads to abandoned applications, reputational damage, and operational inefficiency, smaller competitors can feel left behind in their fraud strategies.

Constantly keeping track of changing patterns enables FIs to analyze their impact on the current strategy. However, understanding the impact could be a very complicated and overly manual process that can potentially lead to bad experiences if not managed well. The key is to find a way to constantly improve fraud strategies, best accomplished with ML.  Each time the strategy is challenged, the better the solution becomes.  So, how exactly is this accomplished?

How machine learning improves fraud strategies

When updating and improving strategies, it’s important to understand which one is the winning strategy. This cannot be fully justified by using sample data or running tests on historical data, as over the period of time the historic and live strategies can differ with the data services they use. If any new data service is used in the live strategy and historic data is used, where this data service was not integrated, then the results will not be reliable.

When strategies need to be improved, one primary concern revolves around marginal cost, and how to keep a frictionless experience for good customers. Efffectiv’s unique capability solves both of these pain points with one feature.

Effectiv’s champion challenger tests two strategies, live & the other strategy before it goes live. The platform’s machine learning model provides a comparative analysis against the current live strategy to determine a winner, the one stopping the most correctly identified fraud while limiting the number of false positives. With each test completed, champion challenger becomes more intelligent, helping stop fraud even better.

How to set up Champion Challenger

Champion challenger allows users to choose the strategy they want to test. Once the user starts the experiment, the strategy being tested will process the data in the live workflow. At any point in time, users can analyze the performance of the live test. Once the performance has been analyzed, the user can stop the experiment and declare the test strategy the “winner” and replace the live strategy – or the user can stop the experiment without declaring a winner. All previous test results can be reviewed at any time.

Here’s an example:

Champion Challenger - Declaring a Winner
Champion Challenger – Declaring a Winner

The metrics available for analysis are comprehensive, including out-of-box metrics provided for the analysis. If needed, users can add or remove metrics based on the kind of analysis they want to perform.

What’s next?

Are you ready to take advantage of Champion Challenger’s ML capabilities to help defeat fraud? Contact us to learn more.

Book a Demo

Get a 30-minute personalized Q&A and demo with one of our solution experts and discuss your specific use cases and questions.

Schedule a call at your most convenient time here.
Or send us a message and we will contact you.