What is alternative data?

We are living in a big data era. Everything is different than what it used to be. A trend I have been witnessing in the past few years is that alternative data has become increasingly popular in the auto finance industry. For instance, when assessing a customer’s creditworthiness, auto lenders will not only rely on the data provided by the three national credit bureaus – Experian, Equifax, and TransUnion, but will also take other non-financial payment data such as utility and telecom payments, transaction data, website usage, even social media posts into account. These non-financial payment data are called “alternative data.”

Why do we need alternative data?

Traditional data can’t fully represent a client’s credit history. There are many creditworthy consumers who lack access to credit. For example, most banks in the United States only offer loans to U.S. citizens and Permanent Residents. What about Non-Residents in the U.S. or millennials that may be new to the workforce who may be financially qualified for loans but are “unscorable”? Auto lenders would be well served to use alternative data in addition to the FICO score in order to target a broader demographic, grow their portfolios faster and concurrently improve loan quality.

There are indeed statistics backing up this point. According to TransUnion’s research in 2015, around 66% of lenders using alternative data say it is helping them “reach more credit worthy consumers in their current markets,” and approximately 56% of lenders using alternative data say “[alternative data] has opened up new markets.” It is worth noting that 53% of automotive lenders were adopting alternative data circa 2015, and the percentage rate has continued to grow.

In addition, auto-lenders may also be well served to consider direct, on-line channels rather than relying exclusively on the indirect channel. Millennials are more apt to researching potential large purchases online such as renting or buying cars. Millennials are more likely to review items online rather than physically go to an auto dealership. In response to this trend, auto lenders should consider using direct, on-line and mobile channels for sourcing applications. Leveraging alternative data becomes more critical for adjudicating loan applications sourced on-line.

How should auto lenders implement alternative data?

I suggest auto lenders think creatively about their credit decisioning processes. Don’t judge a customer solely by his/her FICO scores; take a look at his/her alternative data and then make a loan approval decision.

Alternative data is typically unstructured and come from disparate sources. Developing sophisticated algorithmic models that are capable to making recommendations based on both traditional and alternative data is important to ensure consistency and speed. The online channel can be effective only if credit decisions can be rendered seconds or minutes as opposed to hours or days.

What is the concern of using alternative data?

Nothing is perfect – this universal rule also applies to alternative data. The biggest challenge for auto lenders so far is to figure out a feasible credit model that can fully take advantage of alternative data. With large, dirty datasets, it is hard to develop an algorithm that precisely applies to every credit decisioning process.

However, auto lenders should not feel intimidated by this barrier – this is something data scientists and artificial intelligence focus on. A good data scientist can clearly articulate the predictive capacity of a given model and the associated risk of using it exclusively. The best approach is for lenders to work with data scientists to develop a decisioning process that leverages model recommendations, human intervention (where appropriate) while minimizing time to decision. A technology partner with a strong data science practice can help in this endeavor. (Note: DecisivEdge recently published an article elaborating on how to select the right technology partner. If you are interested, please check out this story for more insights.)

While using alternative data can be very effective in both targeting a larger group of borrowers and in improving loan quality, its use has raised some ethical and legal questions from time to time. While this data is typically sourced legally from public social media posts and other such sources, privacy concerns tends to spook people when these data sources are used in the process.

The best way to tackle this dilemma is to be completely honest with your customers and to make sure that data is sourced exclusively from the public domain.

Conclusion

The journey for auto lenders to employ alternative data into daily practices is arduous and time-consuming, but it isn’t impossible. If we keep exploring the pros and cons of alternative data, we will eventually discover the real value in the credit model.