Learn / Portfolio

Roll rate analysis: how to read a roll-rate matrix

6 min read · Updated 2 November 2026

A roll rate is the share of loans in one bucket this month that are in another bucket next month. Put every bucket against every other and you get a matrix. It answers a simple question: are loans getting better or worse, and how fast?

How the matrix is built

Take the same loans at two month-ends. Rows are the bucket last month, columns are the bucket this month. Each cell is the share of that row that landed in that column. Each row adds up to 100%, once you include a 'closed' column for loans that paid off.

The 25% in the last line is the flow rate from 1-30 to 31-60. The 45% is the cure rate.

Why it is useful

Flow rates multiply. Suppose, for illustration, the rates are: Current to 1-30 is 4%, 1-30 to 31-60 is 25%, 31-60 to 61-90 is 50%, and 61-90 to 90+ is 70%. A loan that is current today ends up in 90+ four months later with probability 0.04 x 0.25 x 0.50 x 0.70 = 0.35%. That is a rough estimate of future bad loans from today's buckets, before any model.

If the 1-30 to 31-60 rate rises from 25% to 30%, you can see the effect on 90+ before it arrives.

Mistakes to avoid

• Using different loan sets in the two months. New loans and closed loans need their own rows or columns.
• Mixing count and balance in one table.
• Reading a single month. Look at six to twelve months to separate a trend from noise.
• Ignoring small buckets. A bucket with 20 loans can swing 10 points on a single loan.

How to read it in practice

• Look at the diagonal. A high diagonal means loans stay where they are.
• Look just above the diagonal. That is the roll-forward risk.
• Look at the cure column. A falling cure rate means collections are losing grip.

Frequently asked questions

Keep reading