How to predict consumer behavior in a post-COVID-19 world
Katerina Folkman August 27, 2020
As 2020 has "blessed" us with a plethora of unexpected challenges, lending institutions have deployed analytics on two main new tasks: (1) anticipate changes in customer payment behavior, to contain portfolio risk, and (2) gain new insights on the future borrowing needs, to ensure quick and safe "Back to Market" return.
The borrowing and payment patterns have changed across all customer and small business segments. The interesting challenge for analytics was to figure out whose risk has truly increased, and which segments have emerged as "new prime", even stronger than before the pandemics. For example, some of the "prime" category A salaried customers have lost their jobs due to lay-offs, while hustling self-employed individuals showed greater ability and willingness to pay, capturing the wave of new customer needs.
Another challenge is that aggregates, averages and historical regression scorecards used by lending institutions before, have lost their predictive power. Historical customer behavior pre-pandemic is not relevant anymore, to predict very rapidly changing new risks and needs. In such an environment, the best bet is new and advanced machine learning/deep learning algorithms that use the most recent, more granular data, for new segmentations and risk scoring.
The addition of alternative customer data is now essential to review customer risk assessment and lifetime value forecast. For example, some leading lending firms have engaged in "customer voice" initiative, connecting with thousands of customers who have requested the moratorium. The data captured during this exercise gave them a new view on customer sentiment and factual behavior, during financial uncertainty.
Analytics help identify new trends e.g. lower income stability, job losses in households, lesser confidence in future income streams, decreased contractibility due to relocations. Additional insights come from digital channels (e.g., online applications or "my account" pages), tracking such granular customer behavior as misspelling of one's own name or hesitations while selecting duration of the loan.
In order to deal with uncertainties, lenders have revised their risk analytics approach, redefining the weights in existing models and adding new moratorium-driven variables. They also define new high-value target segments, looking into changing future customer lifetime value.
Interestingly, the borrowing needs of the customers have changed too. Due to uncertainty about the financial future, fewer customers are interested in traditional rigid lending products e.g. 3-year personal term loans. However, many more expressed a new need for flexible digital instruments. Such customers might be less committed to fixed monthly EMIs, but need more "rainy day" options like credit lines.
The new trend of "mindful spending" has certainly brought down discretionary spending, but at the same time, increased the needs for "new essentials".
Homebody health-conscious customers spend more on broadband, home renovations to make home offices, preventive care, nutrition and vitamins, among other new essential categories. This helps lenders sharpen the new "back to market" approach for business loans, targeting small business customers in thriving industry clusters. It also helps find new low risk - high value customer borrowers, who have new and different borrowing needs.
"Maintain hope and focus on growth" is a new motto for many of us in the second half of 2020, in both personal and professional lives. While aggregate macroeconomic growth is low, we believe it is the best time to grow in a differentiated manner. Companies, in order to position themselves for such growth, need to react faster and continuously track a "new high risk" in traditional segments.
Additionally, the focus should be on new emerging "new high value" customers whose earnings and businesses are only going to grow, even in this new post-pandemic era.
(The author is Head - Analytics, Clix Capital)