A servicer pulls the aging report every Monday and sees delinquency sitting at 8%. Aging reports don’t indicate whether that 8% is getting worse or better, and that gap is why portfolios that look stable on paper still end up surprising everyone with a spike in charge-offs three months later.
The metric that closes this gap is delinquency bucket migration: the movement of a delinquent balance between aging stages over time, tracked forward through roll rate, tracked in reverse through cure rate, with recovery rate measuring the financial outcome of that movement.
In working with private lenders, CDFIs, municipal lenders, and commercial shops across more than seven years of loan servicing software implementation, I’ve watched two teams read the same aging report and reach opposite conclusions. One team calls it a warning sign. The other calls it normal seasonal drift. The disagreement always traces back to the migration data, which neither team had in front of them.
Cason Rentals, LLC cut loan defaults by 20% after replacing static delinquency counts with migration tracking as the primary early-warning signal. That result came from watching direction, not size alone.
Three metrics turn migration into something a servicer can act on before recovery becomes damage control: roll rate, cure rate, and recovery rate. Let me walk you through each.
Roll rate is the percentage of a delinquent balance that moves from one aging bucket into the next, most commonly from 30 to 60 days past due. Divide the balance that advanced into the next bucket by the total balance that started the period in the prior bucket, then multiply by 100.
Roll Rate (30 to 60 DPD) equals the balance that moved from 30 to 60 DPD, divided by the total balance in 30 DPD at the start of the period, times 100.
Most servicers calculate this by hand. They pull two aging reports a month apart, drop them into a spreadsheet, and match loan IDs to see which balances moved forward. Any loan that gets modified, paid off, or renumbered mid-cycle breaks the match, and the roll rate that comes out the other end understates how much of the loan portfolio is actually deteriorating.
Roll rate matters beyond internal reporting. Private credit investors and asset-backed security analysts track roll rate as a dynamic read on portfolio health, since it shows the velocity of deterioration in a way a static delinquency snapshot cannot. A roll rate that climbs from current to 60-plus days past due tells a very different story than a roll rate that stays flat, even when both portfolios show the same overall delinquency percentage.
Roll rate benchmarks aren’t universal. Loan type, borrower credit profile, and economic conditions all shift what counts as a normal roll rate for a given book, which is one reason a single key consumer lending KPI rarely tells the full story on its own.
Bucket-to-bucket reconciliation holds up whether it happens in a spreadsheet or inside a capable loan management system built to track it.
In Bryt, each loan’s DPD (Days Past Due) bucket updates with every posted payment, so a loan that’s modified or paid off mid-cycle stays matched instead of falling out of the roll rate calculation.
Cure rate is the percentage of delinquent balances that return to current status within a given period, calculated by dividing the cured balance by the total balance that was delinquent at the start of the period. A high cure rate means most of yesterday’s delinquent loans need a light-touch reminder, not a hard collections escalation.
Without cure rate data, a collections team treats every 30-days-past-due loan the same, working the list in whatever order it was pulled. A loan with a history of curing on its own gets the same urgent call as one that has rolled forward every month for a year, wasting time on borrowers who were never going to cure without intervention.
Cure rate fixes that. A loan with a strong cure history moves toward a lighter-touch reminder, while a loan with a falling cure rate moves to the top of the list. Loans that cure quickly rarely progress to default, and a declining cure rate across a segment is often the earliest sign that a collections strategy needs to change.
Cure-rate segmentation works in a spreadsheet or inside a capable loan management system that tags cure history at the loan level.
In Bryt, each loan carries its own delinquency and cure history inside the loan record, so a collections team can filter by cure pattern without rebuilding the list every cycle.
Recovery rate is the percentage of a delinquent or charged-off balance a lender actually collects, calculated by dividing the total amount recovered by the total balance at the time of default or charge-off. Lenders distort recovery rate most often by mixing principal-only recoveries with recoveries that include fees and interest, then comparing that blended number against a prior period that used a different formula.
The denominator causes as much distortion as the numerator. A portfolio manager calculating recovery rate against gross charged-off balance gets a different number than a collections lead calculating it against net balance after collection costs, and reporting both side by side treats them as the same number.
A non-performing loan’s recovery rate also depends on when the clock starts. Measuring recovery from charge-off produces a different rate than measuring from 90 days past due, and changing that starting point mid-year makes performance look improved when only the measurement is moved. Cutter Hill Capital improved audit readiness after standardizing its recovery-rate definitions across its servicing team.
A standard recovery-rate formula holds whether it’s written into a shared spreadsheet template or built into a capable loan management system.
In Bryt, recovered principal, fees, and interest post to a loan as separate line items, so the recovery rate is calculated the same way every time without a team reconciling definitions first.
CDFIs and non-profit lenders often report to funders quarterly or annually, while roll rate, cure rate, and recovery rate need frequent, consistent bucket-level data to mean anything. That mismatch is why many CDFIs end up reporting delinquency rate alone, the one number available without extra work.
Federal Reserve Bank of San Francisco research on CDFI financial data found that CDFI loan funds blend public, private, and philanthropic capital, giving each one a structure that doesn’t map onto the standardized metrics banks or fintech lenders use. Comparing one CDFI’s numbers to another’s without accounting for that ends up being misleading.
Municipal and microfinance lenders face the same gap. Payment waterfall mistakes further up the chain corrupt this data before a CDFI reaches the funder report stage.
Quarterly funder reporting doesn’t require daily migration tracking underneath it, only a system that captures the data continuously so quarterly numbers get pulled, not rebuilt.
In Bryt, aging report snapshots accumulate with every servicing cycle, so a CDFI can pull a quarter’s roll rate, cure rate, and recovery rate from existing data instead of reconstructing it by hand before a deadline.
Roll rate, cure rate, and recovery rate answer three different questions: how fast delinquency is spreading, how much resolves on its own, and how much money comes back once it doesn’t. Together, they catch what a single delinquency percentage or portfolio-at-risk figure can miss alone.
National credit card and auto loan delinquency rates flattened through most of 2025, then auto loan delinquency ticked back up in the third quarter, a reminder that an aggregate number can hold steady right up until it doesn’t. [Source]
A lender watching roll rate and cure rate underneath it would have seen the shift building first.
Worcester Financial LLC scaled from 50 to more than 100 loans in two years while cutting operational costs by up to 50%, a pace that depends on catching migration early, not reacting after the fact.
See how Bryt calculates all three from the same payment and aging data, so a servicer reads them side by side instead of reconciling three exports built on different assumptions.
© 2026 Bryt Software LLC. All Rights Reserved.