Confessions of a Data Analyst

The implementation of CECL has been called the biggest change in financial institution accounting . . . ever. Under current U.S. GAAP, financial institutions account for losses based on historical events or incurred losses. Beginning in the first quarter of 2020, financial institutions must look at the past as well as the future over the full lifetime of a loan.

With this change in loss reserve estimation, data is now more important than ever. As a data analyst for MST Advisory Services, I’ve seen the good, the bad and the ugly. The benefit of that is I can let you in on four of the most common data challenges you might face and how best to overcome them. That’s my confession.   

First and foremost, why is data so important in the CECL model? Building a CECL model combines art and science, similar to building a perfect house. You need the math and science behind putting together a blueprint, so the house can stand upright on a solid foundation. You also need a keen artistic eye to make sure the interior is properly decorated. You can’t have one without the other and get a great model. (Here’s another confession, we want you to have a great model. One that’s tailored to your specific financial institution.) 

Without further ado, here are four of the most common data challenges under CECL and some solutions: 

#1 Disaggregated Sources

Disaggregated sources refers to having multiple sources that contain data or information that could be used for CECL implementation. Because those sources are not consolidated, it can be time-consuming to parse through the files and account for all of the data fields available to you. Sometimes these files are not maintained or retained consistently across the board. This can lead to a disparity in the quality of data and how much historical availability an institution has. Another issue we have seen is that the unique identifiers (loan number, account number, etc.) between two sources of loan level data are not consistent and lead to issues attempting to match loans across sources. 

To counteract this, we strongly suggest investing in or building a data warehouse to store all of your data. This eliminates the headache of parsing through sources and is a huge help in maintaining the volume of data you will inevitably collect for CECL moving forward. In the meantime, take stock of your data – what information do you already have? What do you still need? Document where your data is located and the earliest point it is reliable, which helps an institution understand how far back their data is usable. Having consistent coding and a formalized data management process will help better collect and sort the data you already have. Additionally, your staff should be working together and communicating across departments. Is your commercial lending staff talking with your mortgage lenders to complete a customer data profile? They should be. 

#2 Maintaining Statistical Relevance in Pooling

Maintaining statistical relevance in pooling is important to your data process. Are you keeping a big enough sample size for your data? If not, it can produce artificially high or low reserves and make it harder to correlate to external factors. The challenge in pooling is balancing appropriate granularity in the portfolio and maintaining statistical relevance. Make sure you document the decisions you did (and did not) make and the justifications behind them. The transition to CECL provides financial institutions with the optimal time to revisit their loan pool segmentation and make changes, if necessary. 

#3 Updating Risk Factors

Risk factors change over the life of a loan, and it’s important for financial institutions to update and change theirs on a consistent basis. If not, you can face an increasing risk of losing accuracy in reserves and limit pooling, methodology and Q-factor options. Certain risk factors such as risk ratings, consumer credit scores, loan-to-values (LTVs) and days past due are a handful of dynamic risk factors you may want to update during the life of a loan. By updating your risk factors, you will get a better loan review, more granular insight into losses and better isolation of risk. However, updating risk factors requires time, money and staff which some institutions just don’t have. Identify the benefits and the costs of updating your risk factors. If the benefits outweigh the costs, consider investing resources towards updating your risk factors consistently.   

#4 Lack of Historical Loan Level Data

It’s highly important for financial institutions to have historical loan level data. A full economic cycle of historical data is recommended. If not, you risk limiting your methodology and pooling options.

If you do not have the data needed for your institution, we suggest utilizing external data – look at peer data, Uniform Bank Performance Reports (UBPRs) and third-party vendors. Do the best you can with what you have and make sure you plan for the future. CECL adoption is just the beginning so assess where you are now and plan for 3, 5 and 10 years down the road. Identify where you want to be in the future and develop a plan of attack to gather the historical data that you are lacking today to continue to refine the allowance estimation in the future. 

If you are still facing data gaps, think of what your ideal methodology and pooling structure are. What data do you need to start gathering? How long will it take to get there? Also, take into account what your auditors will expect. Remember to include your auditors in this process, too. The sooner the better so they can help guide you to keep you in compliance. They don’t want to be surprised on an audit any more than you do. 

In transitioning to CECL, consider all available information – internal, external or both – “relating to past events, current conditions and reasonable and supportable forecasts.” (ASU 2016-13 Subtopic 326-20-30-7) Financial institutions are not required to exercise “undue cost and effort” in the search for data. You may discover that using internal information works best for your institution. In the end, apply judgement to develop estimation practices that are appropriate and practical for your circumstances and document, document, and document some more. 

If you’re still having issues with data in preparation for CECL, reach out to our team of advisors and we will work with you to make sure you are on track and in compliance ahead of the implementation deadline.

Find out more about the Advisory Team. To schedule a call with an advisory, click on the button below.

About the Author

Zach Langley is a Data Analyst for MST’s Advisory Services and has assisted many institutions in the creation of their CECL Blueprint through Data Assessments and Gap Analyses. He has worked closely with the team of expert managers within MST Advisory and his experience with institutions ranges from $1 billion credit unions to banks with asset sizes over $20 billion. In addition to helping institutions develop their CECL Blueprint, he has also assisted in validating current incurred loss models.

Zach graduated as co-valedictorian with a Bachelor of Arts in Business Management from Piedmont College.

This blog was adapted from Zach’s workshop presentation at the 2018 MST National CECL Conference. Watch for the full digest from the conference this summer.