Content Library

Cutting through the noise with data

Cutting through noise with data

Synthetic fraud? Biometrics? An abundance of jargon and buzzwords emerge when you begin to research identity validation and data in concert with the creation of a digital account opening strategy. This article attempts to go back to basics and outline some key considerations with data when a financial institution turns toward digital channels for growth.

There is a lot of data out there. More data than anyone can wrap their head around and more providers of identity verification this year than last, which makes it challenging to figure out what combination of sources is best for a financial institution.

The approach we’ve seen success in cutting through the noise is closer to science than art. Institutions should be clear about goals, think broad, test (and test again), and experiment with new strategies for identity verification over time. The last point is most relevant when squaring up against an ever-changing fraud environment.

Let’s go through each of the points above. Starting with goals, a financial institution should balance the following:

  • Fraud prevention

  • Automation (benefits in operational efficiency and customer experience)

  • Increased approvals

  • Cost

Every strategy should balance those four categories with the realities of their business. Short-handed? Emphasize automation. Looking to capture more market share? Focus on approvals.

The next question is, of course, “how to get it done?” A strategy we’ve seen work is to go broad with data and set up a program to test and refine your approach. As an example, Alloy performed a large scale fraud prevention study with a major US retail bank. In the test, we looked at the efficacy of different data types in detecting known fraud outcomes. Data types included traditional credit bureaus, device data, digital and social data, and other identity products.


Each data type held advantages in coverage and accuracy with fraud detection ranging between 23 percent and 67 percent with between five and 10 percent false positives. The best solution for fraud detection was to go wide. Including several types of data increased detection rates to 90 percent and decreased false positives to two percent.

It takes work and constant testing, but results can be significant. This content was originally presented at an event hosted by Blenheim Chalcot where thought leaders from tech, law, and finance joined to discuss topics on data and privacy. To learn more visit

Related content