Using Data Modeling to Predict Member Struggles

The COVID-19 recession officially started in February and the economic effects are still reverberating throughout the nation. This recession, with its peak unemployment at 14.7%, has already began to cut deep into your community and your members’ economic health.

This could easily be a time of panic for a credit union leader, but for those using data, this recession and its painful effects on your membership can be dulled. Data can turn a credit union’s products into a community resource rather than another economic hurdle for its members. This article will explore how a credit union leader can use data to predict member struggles so they may be addressed proactively, while also easing enterprise risk.

Leadership is the catalyst for how a credit union deals with this recession and its impacts on members. A leader must be declarative about the vision that will guide the credit union in its approach. Attempting to solve all problems at once will result in tepid and unmistakably disjointed results. Instead, a leader must first observe and then decisively orient how the organization will address the challenge. This is critical to identifying what data will be needed in the models.

There are some data points that are useful in nearly all data modeling exercises focused on identifying and relieving member distress. The first and most telling data point is direct deposit trends. A member’s direct deposit information says a lot about their financial health. An abrupt reduction in amount, change in source or altogether halting of a member’s direct deposit is a strong indication of a distressed member. However, this alone does not answer whether a member is struggling, as they could have simply changed their anchor bank.

To balance out the model, there are two other general data points a credit union leader should consider. First is the rate of late payments, monthly payments that total closer to the minimum payment, or the utilization of payment deferrals on existing loans. A financially distressed member is likely to change their repayment behavior, which could manifest itself as payment delinquency. This is called a strong signal, but likely a latent one. Because it is latent, it is of lower value for a credit union intent on intercepting struggles before they become damaging.

For the model, evaluating external payment behavior is likely a more useful indication of a struggling member. To infer this information, a credit union may look at a member’s historic and current payments to other consumer lending products, like credit cards. If a member regularly makes a payment to a credit card between $500 and $800 per month, and suddenly reduces it to $39 per month, this becomes a signal of projected struggles. Members know their circumstances better than a credit union leader and these types of behavior can reveal future intent.

Another strong data point is luxury spending. A member anticipating a job loss is likely to cut back on discretionary expenditures, and this data can readily be found in credit card, debit card and ACH transactions. Money directed toward non-essential merchants is not all – a credit union should also be including debt servicing, such as on a consumer credit card, in their discretionary spending analysis.

Putting these three data points together can provide a credit union leader with a quantifiable and repeatable model of members who may be struggling when overlaid with corresponding member segments. Once this model is applied to the membership, a leader can get an appraisal of the overall health of the membership and monitor changes. This actionable insight will undoubtedly feed into how the credit union will handle situations that increase enterprise risk and explore options on how to best serve its members. The data may spur the credit union to head in a direction different from its initial vision, or affirm a decision to stay on course. Now the leadership has collective awareness and can make adjustments for the benefit of the member, but not at the expense of the membership.

Moving from general to specific use cases, a credit union leader can take this general model and use it to answer specific questions. If the leadership is interested in understanding whether members are struggling with the credit union’s issued credit cards, the answer can be modeled. The previously outlined general model can roughly quantify the total population of vulnerable members. That now becomes an input for a new model of “what if” analysis. If the leadership decides to offer balance transfers and debt consolidation loans to these members, this model will inform what the loss in fees and increase in loan servicing will be. This is critical for future planning and balancing risk.

This same approach can be used for auto loans. Using the general model to identify which members and how many, the leadership can learn with some certainty what the impact would be of offering waived fees on skip-a-pay or serial skip-a-pay users. The model allows leadership to explore even more creative ways of assisting struggling members since the risk is quantified within a range of reasonable confidence.

The only thing certain about the future is that it is uncertain. However, the Great Recession may be instructive of what credit union leaders might expect from the COVID-19 Recession. The last recession was recent enough that general patterns are likely to still be of use. Using the general model of struggling members against what the credit union learned during the last recession will allow for modeling of enterprise risk.

Comparing the number of member defaults or write-offs from the last recession, adjusted for population change, can serve as a baseline to compare against the new general model. If the Great Recession write-offs adjusted for population change is similar to the member count from the new struggling member model, this suggests that the credit union may be in for similar economic headwinds. This approach of historic versus new models allows the leadership to execute lessons learned from the last recession and preventative measures for the COVID-19 Recession.

As a community financial institution, a credit union is uniquely positioned to address the needs of its members. Identifying those members who are struggling using data and modeling with a preemptive offer of debt consolidation or some other expression of relief will build loyalty that simply cannot be bought. It is not clear how long or deep this recession will be and some of your members will struggle.

Data and modeling can position a credit union in a more favorable position in the future. A credit union leader who uses data gains the initiative; those who do not resign to react.

Ray K. Ragan, PMP is the co-founder of Clear Core, a data cleaning and transformation provider focusing on increasing the value and accessibility of data for financial institutions, in Tucson, Ariz.

Timothy “Buck” Strasser is the founder of Clear Core in Tucson, Ariz.