Improved Targeting for Loans - Mike Jacobs



Questions from the Audience


Question 1: How can you differentiate between credit usage levels among lower-engaged members?

Answer:This is a tricky area, as we would need to define “engagement”. Often, a borrower does not have to be engaged to be valuable; for example, a member receives a $30,000 automobile loan and makes monthly payments via debit order. I would not regard him as being engaged, yet is a valuable member. The Credit Union would need to analyze the relationship between engagement, ownership and value for different product categories.

Question 2: What went into determining the Asset sizes? and what are the implications of expanding it with regards to have the same insights effects?

Answer:The asset breaks are not fixed, but are determined for each specific implementation, to be most appropriate for the particular market. We could have more or fewer asset breaks. More asset classes give more granularity, hence more precise insights -- but this is balanced with added complexity, which will make the results more difficult to interpret.

Question 3: Can the risk model score be used instead of a credit score?

Answer:The risk model score should not be used for any credit decision. The score identifies members or prospects with higher risk of adverse performance, so the Credit Union can decide how to treat these households, communicate with them and manage expectations. The risk score is based on overall financial behaviors and measures probability. A credit score uses detailed credit histories of individuals and thus is generally better for an actual credit decision.

Question 4: Is there any trend data available? How have the statistics changed over time?

Answer:It is very difficult to reconstruct historic positions, so when this system is implemented, we generally have to start with a single point snapshot. Over time, it will be beneficial to monitor changes and establish trends on which segments are growing/shrinking or performing better / worse. But, as stated, we have to start without trend data, unless it is possible to obtain good historic data, which is challenging.

E-mail me when people leave their comments –

You need to be a member of CULytics Community to add comments!

Join CULytics Community