Credit analysis, Level 1: loan portfolio health check

Level 1 of a deeper credit analysis, built in Power BI on 5,000 loan records. It is a preview of the health of the loan portfolio. Later levels, up to Level 3, will go deeper.

A lender needs a quick, honest read of its loan book: how much risk is at hand, where that risk is concentrated, and which segment needs the most attention.

• Give a preview of the health status of the loan portfolio.
• Show how much risk is at hand.
• Show where more of the risk is concentrated.
• Identify the segment that needs more attention.

5,000 loan records covering product type, channel, state, loan amount, debt-to-income ratio, credit score and past-due days. The data is dummy data, used to stay clear of any confidentiality issue.

• What exposure is the portfolio carrying, and how does it compare with what was originally disbursed?
• How much of the portfolio is delinquent, and how severe is it (PAR30 against PAR90)?
• Which product types, states and channels carry the highest risk, not just the highest volume?
• Is the credit score reliable enough to predict risk the way it is supposed to?

An interactive Power BI report. It shows total loans, total exposure, default rate, PAR30, PAR90 and NPL rate, with exposure by product type and by state, NPL rate by product type and NPL rate by credit score band. State buttons and filters for product type, channel, tenor and customer type let you slice every figure.

• Outstanding exposure is ₦8.2bn, out of ₦12.64bn originally disbursed.
• PAR30 is 19.16%, very high against a healthy level of below 5%.
• PAR90 is 3.24%, which is still healthy.
• Loans are still being recovered before they reach non-performing status, but a large amount of money sits in early delinquency (PAR30). That pool is the pipeline that feeds PAR90 and future non-performing loans, so it needs close monitoring.
• BNPL carries the highest PAR30 at 20.6%, with the most loans unpaid beyond 30 days.
• Asset Finance has the highest NPL rate at 3.95%. These are two different risks, and looking at only one measure would miss the other.
• The NPL rate by credit score band gives a mixed picture, which questions how reliable the credit score is as a predictor of risk.

Finding numbers that do not make sense, and asking further critical questions, is what makes an analysis worth doing. I almost did not share this project because a voice kept saying it was not good enough. Then I remembered it is only the first level of a deeper credit analysis. Level 2 is in progress.