Skip to main content
Secret to Scoring More Consumers? Alternative Data—With A Catch

As I discussed in my last blog post, many of our banking clients have expressed interest in scoring more US consumers. But we've heard little support for doing so by loosening the minimum score criteria used by the FICO® Score and relying solely on traditional credit reports. Clearly, lenders are wary of making credit decisions based on extremely marginal or stale credit bureau files.

There is a better way—one that our research has shown can deliver statistically meaningful credit scores for 60-75% of traditionally unscorable consumers. The secret? Supplementing traditional credit bureau data with alternative data.

Before rushing to declare alternative data as a panacea, the devil remains in the details. Not all alternative data passes muster in terms of driving a more effective credit risk score.

In today’s Big Data world, there are many alternative data providers and thus many potential new data sources. Having more data is generally good, but it is important to resist the temptation of indiscriminately throwing multiple data sources together to bake up an alternative data credit score. This development process needs to be done thoughtfully, with the same meticulous consideration that goes into building the FICO® Score, to navigate through the challenges of today’s regulatory environment.

It’s important to make the right call on which alternative data you leverage, especially given the potentially significant operational and cost considerations of acquiring, maintaining and updating that data. Based on our expertise, alternative data sources are most useful in credit scoring if they demonstrate:

  • Regulatory compliance. The data source must comply with all regulations governing consumer credit evaluation. To comply with the Fair Credit Reporting Act, for example, a data provider must have a process in place for investigating and resolving consumer disputes in a timely manner.
  • Depth of information. The deeper and broader the data, the greater the value. Consider a utility data repository. Does the data reflect both on-time and late payments? Is the account history captured from the beginning of service or just for a recent period? If the consumer has moved, are there records from multiple addresses?
  • Scope of coverage. A database covering a broad percentage of consumers is optimal. If 40% of US adults live in rentals, then a random list of consumers matched against a rental payment database should yield a 40% hit rate.
  • Accuracy. Clearly, inaccurate data compromises the value of the data. Alternative data repositories need to have a mature data management process in place to ensure data accuracy. It’s important to ask questions like: How reliable is the data? How is it reported? Is it self-reported? Are there verification processes in place?
  • Predictiveness. The data should predict future consumer repayment behavior. For example, analysis of a non-sufficient funds (NSF) check database shows that consumers with no bad checks on record are more likely to pay their credit obligations than those who’ve written one or more bad checks. Such a data source would add value for credit risk evaluation.
  • Orthogonality. Useful data sources should be supplemental or complementary to what's captured in traditional credit reports. For example, if a repository collects foreclosure data from public record information, that data may provide little added value to foreclosure data already in consumer credit reports.

The use of alternative data in scoring has become a hot-button issue as of late, even capturing the attention of leaders on Capitol Hill. We’ve been working with alternative data for many years. Our research has found several types of data that are especially predictive of future repayment behavior, among them: deposit account information, supplemental public record information, utility payment history and property/asset data.

Blog_Graphic_Figure_4_450-px
We’ve built scoring models using alternative data and validated them on client portfolios across numerous industries. The chart above shows validation results in three key banking industries: credit card, auto and mortgage. By utilizing alternative data, this FICO-developed credit risk score effectively rank-orders risk (i.e., the higher the score, the lower the associated credit risk) for 60-75% of previously unscorable consumers.

Since there's been a lot of interest in our research on scoring more consumers, I'd like to once again shamelessly plug our Insights white paper on this issue: To Score or Not to Score? (login required).

related posts