ANALYTICS MODEL DEVELOPMENT has come a long way in a relatively short period. What used to be a little understood process managed by data scientists is moving into the mainstream, thanks to the proliferation of big data and end-to-end platform technologies. Equifax fueled this market shift with the 2017 introduction of Equifax Ignite®, a cloud-based data and analytics platform. Ignite has enabled data analysts and business users alike to readily collaborate with data scientists to build the right models that drive the right outcomes.
Today, Equifax’s customer-first strategy is leading the way for new innovations and new opportunities for business growth. It is also a key driver for the company’s leadership in differentiated and alternative data. John Fenstermaker, Equifax’s chief innovation architect (pictured below), spoke about using the right mix of data and modeling techniques to recommend the highest-performing approaches to support smart decisioning and relevant services for businesses to help consumers live their financial best.
TELL US WHAT DIFFERENTIATES EQUIFAX’S ANALYTICS MODELING.
JOHN: We start off by reframing the development mindset. With a customer-first approach to model development, it seems the first question might be, “What do you want to build?” At least, that is how many providers approach model development. We take a different approach一one that proactively shows our customers, “Here’s what we know works best, based on your unique needs.”
First, we gather a customer’s trade and inquiry data. Then, we run hundreds, if not thousands, of simulations, each time tweaking and adjusting the data components as we go. If the customer’s inquiry data is not available, we build a “like database” of peer inquiry groups. All data is aggregated and anonymized during testing.
Throughout this process, we experiment and layer on expanded data sources—including several alternative and consumer-contributed data sources that our customers may not even know exist—to optimize outcomes.
We are looking for the exact combination and calibration of data inputs that can either 1) increase account approvals while keeping risk levels steady, or 2) lower risk levels, while keeping approvals steady.
Once we figure out the ‘secret sauce,’ we go back to the customer with our top recommendations. This is all done prior to model development to preserve the customer’s time and resources, and ultimately deliver the highest-performing model possible, in the shortest time frame.
DISCUSS HOW YOUR TEAM DERIVES EVEN MORE VALUE FROM DATA.
JOHN: This creative approach—tinkering with anonymized alternative data sources and layering it on top of a customer’s existing data—enables a better understanding of individual consumers, and as a result, uncovers hidden opportunities and risk. In turn, these deeper insights allow the business to make better, more personalized credit offers to a wider audience of consumers, including those who might be near-prime, sub-prime or unscorable.
Alternative data is essential to this approach for a couple of reasons. First, it is typically not included in a traditional credit report, which means it offers another angle or view of the customer. Second, it can be highly predictive of future account performance by revealing a consumer’s payment behaviors, credit capacity and credit trajectory.
Here are a few examples of the types of alternative data we use in our simulations:
• Consumer-permissioned data such as bank transaction or telecommunications, pay TV and utility payment data, which show how people pay their “everyday bills”
• Employment and income data that indicates how long a person has worked for an employer, which demonstrates employment stability
• Alternative lending data from short-term, unsecured lenders (such as payday lenders and cash advancers), which can show if a person is making payments on time
Just as important as our data strategy is our ability to experiment in the cloud using Equifax Ignite. It ultimately enables this “plug and play” approach to simulating and building models on an entire data population, as opposed to a sample set that data scientists have historically used.
To read more about Equifax Data & Analytics innovations, click here.