The Rise of the Attributes
It started subtly, but has been gaining momentum in recent months. Attributes, once obscure and little understood, are becoming more common and integral in lending processes. If you’ve heard people talk about attributes, but never really understood what they meant, this column may provide the insight you need to understand how attributes are changing the face of credit.
Allow me to start with a definition, “Attributes 101” style. Attributes are the variables used in credit policy to create credit scores. They are derived from raw data to present a consumer’s credit profile in a more tangible and actionable way. For example, a common attribute is “30 Days Past Due”. This attribute is a summary of how many accounts in a consumer’s credit report have been marked as 30 days past due within a predefined time frame. In essence, the attribute summarizes the raw data in a usable form. While lenders may calculate hundreds of attributes on a credit report, most credit scores use only a dozen or so.
Over the past few years as more and more institutions have incorporated alternative data, most have relied on calculated scores and a simple table to combine scores, for example a FICO score combined with a RiskView score can provide more accurate rank ordering of risk. They paid little mind to the underlying attributes. With increasing competition for customers driving the need for more accurate credit risk decisions, more institutions are seeking to improve their scoring models. To do this, they need access to the underlying attributes in their scoring models.
While credit bureaus have long offered access to both standardized and custom attributes, today we are seeing an increasing number of alternative data providers offering attribute level access to their repositories. In fact, two leading providers, ID Analytics and LexisNexis both recently announced the availability of the attributes behind their scores. This presents a tremendous opportunity for institutions with capacity in their risk organizations to evaluate the efficacy of these attributes in predicting credit risk. Although it has been well established that alternative data can provide significant additional insight, lifting approval rates by 15%, many sophisticated risk organizations have been reluctant to incorporate scores without the ability to evaluate the underlying attributes.
As a result of this increased access to attributes, we expect to see new models based on alternative data soon being implemented at the nation’s largest lenders. The result will be an increased level of accuracy in credit risk scorecards and a subsequent improvement in lending profitability. As this transition occurs, the pressure on smaller institutions to follow suit, improving the accuracy of their models, will mount quickly. Look for leading providers, like FICO, to release a steady stream of more powerful models and capabilities to meet this demand in the coming years.
The one area where attributes offer the greatest unrealized hope is big data. Much has been written about the potential opportunity to utilize the treasure trove of information stored within big data, yet stories of victory are few. Unfortunately, accessing that data in unstructured formats across multiple repositories has proven challenging for even the most sophisticated lenders. The institutions who are able to reduce the complexity of big data into a series of predictive attributes will lead the next generation of credit risk models. For example, consider an attribute that describes a prospect’s profitability. Or a “socially connected” attribute that describes the level of social network conductivity an individual possesses. While some will debate the credit risk value of this attribute, it certainly has significant value for evaluating fraud risk during the account opening process.
The pressure on data repositories to share their information in the form of attributes is undeniable. As this information becomes more accessible in the coming years and incorporates a broader range of data, the potential for financial institutions to make radically better decisions is tremendous. And not just for credit risk, but better decisions across the entire customer relationship. The one question still unanswered is if there will be enough statisticians to take advantage of the new opportunities being presented. If your kids are heading to college soon, you might encourage them to take more stats.
