Operations2 min read
Roll rates: the number that tells you about next quarter
Recovery percentage tells you how last month went. Roll rate tells you what is coming. It is the closest thing collections has to a leading indicator, and it is usually calculated wrong.
A roll rate is the share of accounts in one bucket at the start of a period that have moved to the next bucket by the end of it. Ten thousand accounts in 1 to 30 at the start of the month, and fifteen hundred of them in 31 to 60 at the end, is a fifteen per cent roll. It is a simple number that most operations either do not compute or compute in a way that hides what they need to see.
Why it beats recovery percentage
Recovery percentage is a ratio of money collected to money due, and it is dominated by the composition of the book. A month with more easy accounts looks like a good month. Roll rate follows the same cohort forward, so it isolates whether the operation actually changed behaviour, and it moves before the recovery number does. That is what makes it worth watching weekly rather than at month end.
Three ways it gets computed wrong
- Counting accounts that entered the bucket mid-period. The cohort has to be fixed at the start or the denominator drifts and the trend is noise.
- Ignoring cures. An account that went from 31 to 60 back to current is a different outcome from one that stayed at 31 to 60, and merging them hides your best result.
- Reporting by value only. Value roll and count roll answer different questions, and a book where the large accounts behave differently from the small ones needs both.
What to do with it once you have it
Segment it. Roll rate by product, by vintage, by originating channel, by city, by whether first contact was made within the first week. That last cut is the one that usually pays for the whole exercise, because it is the first honest measure of whether early contact is doing anything. If accounts contacted in week one roll at the same rate as accounts contacted in week three, your early-bucket effort is decorative.
Backward roll deserves a name too
Cure rate, the share of a bucket that improves rather than worsens, is the mirror of roll and is reported far less often. An operation optimising only against forward roll will happily hold accounts static. One that watches cures is measuring whether it is actually resolving anything.
A caution
Roll rates respond to changes in policy as much as to changes in performance. A new restructuring scheme, a change in what counts as a cure, or a write-off run will all move the number without anyone having collected differently. Annotate the series with what changed, or somebody will eventually present a policy artefact as an achievement.