Starting with an end in mind makes the process of using data to make better decisions simpler and more effective.
If you’ve read enough of my columns, you should have a sense that I write about things I’m experiencing in the office. That’s intentional! First of all, it makes the writing process a lot easier! Secondly, I think my process results in content that’s much more interesting and adds value for the reader.
Collecting and utilizing data to generate value for your organization should have a process as well. Before tracking down data, you should have a business need and an idea of what you want to do with the data. Starting with an end in mind makes the process of using data to make better decisions simpler and more effective.
My Recent Experience
Recently I had some one-on-one time with an analyst in my lending group. She had compiled some data on our auto repossession history and was working on making heads or tails of it. She showed it to me to get some feedback and, frankly, there was no way I could analyze the data and make better business decisions from it.
First of all, the data was too broad, encompassing data from loans made back to 2004. Secondly, due to the wide span of the data, 17 years, the growth in lending horribly distorted the data. Between 2004 and 2021, the dollar amount of car loans we made grew almost 10-fold. We picked up a lot of market share (almost triple in our core market), plus the size of the average car loan had increased dramatically in that time. The early years in that data showed a very low proportion of our overall repossessions. No surprise to me. In comparison, the years where we really “turned on the jets,” 2014-2018, showed a disproportional number of our repossessions.
If we didn’t have a way to truly understand this data, we could use it to make some really bad decisions.
The Root Problem
As I explained to “Chris,” the mistake many managers make is not having a plan for their data—or having a plan that’s backward. What I mean by that is that managers seek out data as a first step. Once they have the data, they figure out how they’ll analyze it, and only then do they start thinking about the business decisions they’ll make with it. From my perspective, this is all wrong, even backward.
To effectively utilize data, you must first have a specific business problem or challenge. Then you should have a feeling of how you might use the data once you have it. That will allow you to search for the right data! Sometimes the business challenge can be a theory. For example, Chris theorized that over the last few years we’ve had a disproportionate amount of our repossessions coming from newer loans.
Chris had the first piece of the process correct—she had a theory.
How Good Is Your Theory?
Chris is one of my most trusted people; she’s worked for me and with me during my entire tenure at Ent Credit Union. When I came here, I brought a lot of new ideas and ways of looking at risk, and she was one of the first people to really embrace what I was hoping to do!
Yet her theory in this case wasn’t that strong. If we had an uptick in repossessions coming from newer loans, we’d probably also see losses rising that were otherwise unexplained. While we certainly have higher delinquency after two years of COVID-related issues, our delinquency is still very low compared to credit unions overall. There were no other warning signs.
How Should We Approach the Analysis?
Chris had a theory and, for now, let’s assume she had a stronger case for believing there might be a problem. Yet she didn’t know how to analyze the data she collected. That resulted in her compiling data that perhaps didn’t hit the mark dictated by her perspective on the potential business challenge.
For Chris to effectively analyze the data, she had to have some understanding of the big picture and other pieces of data on the periphery, so to speak. What did she understand about repossession data? I reminded her of things we did know. For example, the weaker the loan is, the more likely it will default earlier in its life than other loans. Also, the older a vehicle is at repossession, the more likely the car loan is to end in a repossession. We also talked about the COVID-19 impact: Repossessions were delayed by the massive number of loan deferments and modifications done in 2020; so to include data on loans that defaulted in 2020 was probably worthless.
So, what data could we use?
At Ent CU, we have data on the time from loan origination to default for every repossession on a monthly basis. That’s good, but that data includes all defaults and doesn’t focus specifically on newer loans.
What if we looked at all repossessions in 2021 and determined the percentage and number of repossessions that occurred within 12 months of origination? That’s a sufficiently narrow scope. Then we could compare it to the repossessions we incurred in 2019 to avoid COVID-19 issues. Notably, the increase in our originations between the two periods had not grown exponentially; if it had done so, it could have distorted either the data or our analysis.
Now, that’s an example of having sufficient understanding of how to analyze the data!
The Right Data
Having a business need or challenge combined with an understanding of the potential decisions you’ll make with your data along with a general feel for how you’d analyze it is critically important. If you’re missing one of these steps in the process, you might have a challenge finding the right data to make the decisions at hand.
CUES member Bill Vogeney is the chief revenue officer and self-professed lending geek at $8.7 billion Ent Credit Union, Colorado Springs.