We have worked with more than 200 credit unions in the last decade and only a few were able to initially provide clean and accurate data. You will need to have at least 24 months of robust data to move to the analysis stage, and if you have only begun to collect some of that data, it's not hard to see that you may already be behind. In some cases, you will need to work with third parties or extract data from archived data storage, all while pursuing your normal course of business. Simply put, this will take time, so you need to start now.

Analysis (12 to 24 Months)

CECL is highly dependent upon your knowledge of how your portfolio performs at a granular level, well beyond simply the credit score. In order to predict how future loans will perform, you have to look back at how loans with diverse parameters have performed differently in the past. How does a 740 borrower with a 50% debt to income compare to a 740 with a 25% debt to income? They are different and you need to know why. How do economic changes affect these two borrowers differently? Studying static pools of loans with different criteria will help you better understand that. Credit unions need to implement tools that go beyond observing static reports to actually analyzing cause and effect.

Once you have clean data, you will begin to slice and dice that data. With the multitude of variables that can affect a loan portfolio's performance, you may go down one path that leads to a dead end and then have to try a different path. Simply put, this takes time. If you have never played the role of analyst, you can fall victim to many of the fallacies associated with analytics and for this reason, it takes time to learn from one's own mistakes.

Predictive Modeling (Six to 12 Months)

Once you understand how loans have performed in your portfolio, you can build models that predict the performance of future loans. This is the key to getting reliable CECL projections. It's not that hard to predict the chances of your favorite team winning a particular game against a particular opponent, for example. That's because you have reliable statistics that can be used in different scenarios to come up with the probability of a win. Modeling loan portfolio performance is no different, but just as an amateur gambler would only be lucky to correctly predict the outcome of the championship game, without having worked with loan portfolio data in the past, it will be difficult to accurately predict loan losses in the future. CU Direct's Advisory Services has been working with credit unions over the last 18 months to begin creating predictive models, and the results have been highly accurate and informative. Having accurate models, based upon your credit union's data, will help you not only conquer the CECL beast, but will help your credit union be a better lender.

For this part of the process, you may need to seek outside assistance. You will want to consider that the world of analytics and data modeling is the fastest growing segment of business technology. Data scientists are in high demand, and if you start the process today, you will most likely need to stand in line for results. This is not an area where an “easy-button” approach will work. One size doesn't fit all – one credit union can't use another credit union's predictive model as their own, so, be leery of “magic” software solutions that promise turn-key results.

Testing (12 to 24 Months)

You will need time to test your models against real world scenarios and make changes to ensure that your game plan is as reliable as it can be.

So, one can see, adding up the cumulative time that it takes to prepare for CECL, you may only have about a 12-month window to safely hesitate before doing something. That is only if you can confidently say that your credit union is good at executing on long-term projects. If not, you need to start this process today.

Michael Cochrum is VP, Analytics & Advisory Services for CU Direct. He can be reached at 972-407-1710 [email protected].