Evaluating fraud solutions

5 Considerations When Evaluating Fraud Solutions

The fraud prevention ecosystem is chock-full of solution providers and it can be a bit dizzying navigating the different technologies, industries and use cases in which each one specializes. Everyone knows that a “layered approach” is best when fighting fraud, but layered wrong and all you wind up with are stacks of inefficiencies and/or holes.

The devil is always in the details, but here are some guiding principles when procuring the best fraud technology for your organization:

Convergence is Key

  • We are seeing many use cases converge: authentication is blending with fraud and fraud with AML. Blending the data and technology that addresses these use cases only makes sense.
  • These centralized locations go by different names depending on the scope (authentication hubs, orchestration hubs, risk hubs, etc); however, the more important aspect is understanding the function they serve —> managing the multiples…
  • Multiple data streams (ie: internal/external) and multiple use case (ie: payments fraud/AML) on the same platform across multiple teams. This is a significant undertaking, but the long-term dividends are well worth it.

Not All Orchestration is Created Equal

  • Orchestration is a popular industry buzz word, as it insinuates an intelligent utilization of data and decisions. It’s important to dig a bit deeper and ensure strong analytics underpin orchestration hubs.
  • Risk-based analytics are the point at which a basic aggregation hub evolves into actual orchestration.
  • The hallmark of true orchestration is leveraging data and workflows in a more effective and efficient manner, not simply pooling data in one place.

Data is King (Sometimes)

  • Generally speaking, “the more data the better” is useful when assessing the need for rich data in a machine learning model. However, there are some important caveats…
  • Some data sources are redundant and provide the same insight and uplift.
  • Additional data sometimes causes more noise to your machine learning model and actually hurts performance.
  • To get the most bang for your buck, leverage experts to help understand the data sources and map them to a well-designed schema for each fraud use case.

Consider Impact to Operations Teams

  • Management teams in operations should have input into how the tools and technology will fit in with their current workflows and processes.
  • New data and technology must be evaluated from all angles and effectively utilized by the folks making the final decisions on the alerts.

Make them Prove their Concept

  • Every POC will look different depending on the solution you are procuring (ie: data enrichment v. fraud platform), but it’s important to obtain quantitative validation for your organization.
  • Case studies are nice, but remember the cases they are studying are for organizations with different technology, data and resources. These results don’t simply map over to your company.
  • A well-structured POC serves as the best tool to understand potential performance and value.
Viewed 171 times / 1 views today
Tagged with
Posted in ,
PJ Rohall
Author: PJ Rohall
PJ Rohall is a Fraud SME with Featurespace, a pioneer in Adaptive Behavioral Analytics. He also is a co-founder of About-Fraud.com.