The recent credit bureau data breach that compromised personal data of millions of consumers reminds us of the value of consumer data and its relevance to consumer lending.
A few years ago, the initial value-proposition that several upcoming online lenders were offering was – improvement in predictive modeling/credit scoring using personal information from social media channels such as Facebook, LinkedIn etc. The idea was that online lenders would apply machine learning techniques to ‘big data from social media’ and this would improve the credit evaluation process. An interesting white paper by PwC on this topic can be found here. While this continues to be work in progress, some of the social media attributes that are being used for credit modeling purposes include age, gender, marital status, and mobility. Some additional data points are also being used to establish a trust / distrust factor, from the perspective of extending credit.
Among other aspects, one of the major challenges of personal, consumer data in social media is that it is mostly self-reported. Also, this data tends to be more descriptive as against predictive. Just like the Myers-Briggs Type Indicator that helps to understand and create context around personality types, but is rarely used to predict future behavior as personality types are evolving and there is no ‘right’ or ‘wrong’ personality type.
One online lender that has taken advantage of this online personality type profiling is Payoff in Southern California. Payoff uses the OCEAN framework to assess a ‘financial personality’ of visitors to its website, through a quick survey and accordingly adapts its customer service to match the preferences of the customer. But how would any of this impact the consumer’s probability of becoming delinquent on their outstanding debts over a specific period of time? Is there a relationship between personality types and credit behavior?
Credit Bureaus, on the other hand, have hundreds of attributes from factual, consumer ‘transactional data’ that has been consistently collected (from various reporting agencies), filtered for errors and cataloged into a monthly time series over decades and can be linked back to one unique identification number, the consumer’s Social Security number. Data from credit bureaus is extensively used for predictive modeling using advanced statistical methods/ models.
Impact of Data Breach to Consumer Lending
Equifax’s recent data breach has resulted in compromising Social Security Numbers and personal information of an estimated 143 million consumers. This population is almost equal to the total working population of the United States! Some recent, comparable data breaches include:
- Anthem Inc reported compromising personal information of 80 million consumers (2015)
- Experian reported data breach that resulted in compromising personal records of 15 million consumers, including customers of T-Mobile (2015)
How is this incident going to impact the online lending community? Some areas of concern include:
Increase in Acquisition Marketing costs
Online lenders have been voraciously consuming credit bureau data (FICO scores, outstanding revolving debt, number of inquiries etc.) for acquisition marketing, so much so that online personal loans, pre-screened, direct mail campaigns are almost catching up to the humongous credit card direct mail volumes.
With the latest data breach, the FTC is recommending several measures to protect consumers from ID theft and fraudulent use of personal data including that consumers freeze their credit. While a credit freeze does not include ‘opting-out’ of pre-screened offers, some consumer protection extremists are also suggesting that consumers opt-out from receiving pre-screened offers, as mail theft is rampant and such credit offers could be intercepted and ‘misused’.
Such a move (consumers opting out of receiving pre-screened offers) may choke the supply of potentially ‘good credit’ prospects for online lenders to market to. This will result in a higher frequency of mail volumes to the same group of people (who prefer not to ‘opt-out’), thereby dampening response rates and increasing marketing cost per funded loan.
Decline in Credit Approval Rates
The source credit bureau for FICO scores at the time of acquisition marketing can be different from the source credit bureau FICO score used for underwriting. This is so because all FICO scores are not made equal; the methodology used for calculating FICO scores varies by Bureau and a consumer that has a 660 FICO in one bureau, say Transunion may not have the same 660 FICO at Equifax. So for instance, if the underwriting cut-off for a loan qualification is a FICO score of 660 and the consumer is selected using say, Transunion’s FICO and later underwritten using Equifax’s FICO (which may be the median FICO in a tri-bureau credit pull) that comes out to be lower than 660, he/ she may not qualify for the loan.
The point is if the availability of prospects shrink and there is more scrambling for multi-bureau sourcing of mailing lists, the incidents of mismatch in FICO scores (and rejection rates) will increase, resulting in higher cost per funded loans.
The impact of this may either be absorbed by the online lender or passed on to the consumer in terms of higher fees.
Increase in Operations Risks and related costs
With millions of social security numbers (and other personal details) floating around on the ‘dark’ web, the probability of fraudulent activity from fake identities is expected to increase. New measures to screen and verify the loan applicants may not only slow down the process but also increase costs of incorporating new tools and measures to manage fraud.
Disclaimer: The views and opinions expressed in this article are solely those of the author. Examples of analysis performed within this article are only examples. They should not be utilized in real-world analytic products as they are based only on very limited and dated open source information.
Leave a Reply