How do you calculate cumulative net loss (CNL) on a loan pool?

Defaults since the cut-off date, less recoveries, over the pool's original principal: what four credit documents fix in writing, how a trigger schedule rises by month, and what Mexican rated pools measure instead.

28 October 20266 min read
Risk45%CNL trigger from month 24, NetCredit 2023

Cumulative net loss is the principal of every loan in a pool that has defaulted since the cut-off date, less what has been recovered on those loans, divided by the principal the pool started with. It is read pool by pool and month by month, and a credit line tests it against a schedule that rises with the pool's age.

Three choices make the formula, and credit documents write all three down

The ratio looks the same in every document: cumulative defaults, less recoveries, over the original balance. The numbers differ because each document fixes three things. The first is when a loan counts as defaulted. The 2023 indenture of NetCredit Combined Receivables, an Enova vehicle, filed as an exhibit to Enova's 10-Q, counts a receivable as charged off once it is «65 or more days past due» or written off under the servicing policy. The second is what is netted: «Net Liquidation Proceeds» there, «all Recoveries» in the Conn's 2023-A notes (Form 8-K, August 2023), what came back after default.

The third is the denominator, and it is the one lenders get wrong. Every document here divides by principal at the start: «the Original Receivable Principal Balance» for NetCredit, the balance «as of the Cut-Off Date» for Conn's, the principal «at the time of origination» of each vintage, the loans originated in the same month, in the credit agreement between Consumer Portfolio Services (CPS) and Citibank. Oportun's 2024 amendment counts only losses after its revolving period ends, over the balance on that date.

Four credit documents, one ratio, three choices fixed in writing

NetCredit Combined Receivables 2023 indenture (Enova, 10-Q, exhibit 4.3); Conn's Receivables Funding 2023-A Series Supplement (8-K, exhibit 4.2); Oportun CL Trust 2023-A, Amendment No. 1 (10-Q, exhibit 10.5); Page Eight Funding (CPS) with Citibank, Eighth Amended and Restated Credit Agreement (10-K, exhibit 10.1.8). All filed with the SEC, read 27 October 2026.

DocumentDefault pointNettedDenominatorWhat the test does
NetCredit, 202365+ days or charged offNet liquidation proceedsOriginal principal sold to the issuerBreach sends all cash to the notes
Conn's 2023-ADefaulted receivableAll recoveriesBalance at the cut-off dateBreach moves principal to the senior class
Oportun, 2024Defaulted receivableRecoveriesBalance when revolving endsTested in amortisation only
CPS and Citibank, 2024Liquidated receivableNet of recoveriesVintage principal at originationProjected CNL feeds the advance rate

The denominator never moves after the start. What changes from deal to deal is the day a loan becomes a loss and the date the clock starts.

Divide by the current balance and the same pool looks twice as bad by month 12

By our computation, take a modelled pool of 10,000,000 in two-year instalment loans, where 10.35% of principal defaults over its life and a quarter of each default comes back within six months. By month 12 the pool has lost 5.75% of its original principal, net of recoveries. Only 44.2% of it is still outstanding, so the same losses over the current balance read 13.01%. By month 18 it is 7.42% against 39.63%. The ratio over the current balance rises because the pool shrinks, not because it performs worse, and no credit document above uses it.

The same losses read 5.75% of the original pool and 13.01% of what is left at month 12

Our computation, a modelled pool of 10,000,000 in 24-month equal-principal loans; defaults shaped to peak in months 6 to 9; recoveries of 25% of defaulted principal, 10% after one month, 10% after three, 5% after six. Percent of original principal unless stated.

Months on bookDefaults, cumulativeRecoveries, cumulativeCNL over originalBalance leftLosses over balance left
Month 30.40%0.01%0.39%87.1%0.45%
Month 62.40%0.20%2.20%72.7%3.03%
Month 94.90%0.67%4.23%58.1%7.28%
Month 127.00%1.24%5.75%44.2%13.01%
Month 189.55%2.12%7.42%18.7%39.63%
Month 2410.35%2.52%7.83%0.0%Not defined

Lifetime CNL is 7.76%, reached after the last recovery. The ratio over the current balance has no ceiling, which is why no credit document tests it.

CNL can fall. In our computation, after the last default, late recoveries keep arriving, and the modelled pool ends at 7.76% after standing at 7.83% in month 24. The ratio is built per pool, one origination cohort at a time, which is why it lives in a static pool table: each row a vintage, each column a month on book.

The trigger is a schedule by month, and it starts with more room than the pool needs

A credit line does not test CNL against one number. NetCredit's indenture, an exhibit to Enova's 10-Q of July 2024, sets a level for every collection period: 5.00% in the first two, 7.50% in the third, 20.00% in the sixth, 32.50% in the twelfth and 45.00% from the twenty-fourth. Conn's 2023-A notes write theirs by calendar month: 4.00% in September 2023, 9.64% in February 2024, 17.33% in August 2024 and 29.60% from July 2025, next to a separate annualised net loss test of 40.00%. Once CNL exceeds the level, NetCredit's notes enter an amortisation event and all remaining collections pay them down; at Conn's, principal goes to the senior class first.

Both schedules allow a third or more of their final loss by month 6; a normal pool has taken 28%

Trigger levels from the NetCredit 2023 indenture (by collection period) and the Conn's 2023-A Series Supplement (by determination date, September 2023 as month 1). Shares of the final level and the modelled pool's share of its lifetime CNL are our computation.

MonthNetCredit levelShare of its final 45.00%Conn's levelShare of its final 29.60%Modelled pool, share of lifetime CNL
Month 15.00%11%4.00%14%0%
Month 37.50%17%5.98%20%5%
Month 620.00%44%9.64%33%28%
Month 1232.50%72%17.33%59%74%
Month 1837.50%83%24.25%82%96%
Month 2445.00%100%29.60%100%100%

The cushion is widest early and narrows towards the end. A pool that breaches in month 6 is losing far faster than the deal was priced for, not running slightly late.

A young pool is projected to its lifetime loss by dividing by a timing curve

Six months of CNL say little until they are set against the curve older pools followed. The 2024 credit agreement between CPS and Citibank, filed with CPS's 10-K, writes the method down: «Projected CNL» is «the quotient resulting from dividing the most current Cumulative Net Loss Rate for such Vintage Pool … by the corresponding point on the applicable timing curve», and the projections are updated «in the Advance Rate Calculator on each Settlement Date». The curve itself is redacted in the filing; the mechanics are not.

In our computation on the modelled pool, older vintages reach 28.3% of their lifetime CNL by month 6 and 74.1% by month 12. A new vintage that shows 2.86% at month 6 projects to 2.86 divided by 0.283, or 10.1% over its life, against 7.76% for the history. The lender sees the deterioration eighteen months before the last loan is due, and so does the fund, through a lower advance rate.

28%

Share of lifetime net loss a modelled two-year pool has taken by month 6

Our computation; the timing curve divides a young pool's CNL to project its lifetime loss.

A Mexican rated history measures gross default by cohort, not CNL

Mexican lenders that have issued on the stock exchange already keep vintage data, but HR Ratings reads it differently. It rates the trust certificates by comparing the Mora Máxima, the share of expected collections a structure survives losing in its stress case, with the Tasa de Incumplimiento Histórica, the historical default rate of the originator's cohorts, and states the ratio as Veces Tasa de Incumplimiento. The historical rate counts loans that fell past due in each cohort, not what was finally lost after recoveries, and the definition changes by deal, from the weighted cumulative default of each cohort at HIR Casa to the weighted average of the three most recent mature cohorts at TIP Auto.

Five rated Mexican pools: historical default of 1.0% to 7.8%, stress cases of 22% to 36%

HR Ratings rating reports and stock exchange notices: CONCRCB 26S (19 August 2026), EXITUCB 26 (26 May 2026), ENGENCB 26 (27 April 2026), HIRCB 25 (24 June 2026), TIPMXCB 23 (22 August 2025). Historical default is the agency's Tasa de Incumplimiento Histórica; stress loss its Mora Máxima; the multiple is one over the other.

Issue and originatorHistorical defaultStress loss survivedMultipleTrigger in the deal
CONCRCB 26S, FinCrementar7.8%36.1%4.7xNot stated in the notice
EXITUCB 26, Exitus Capital5.5%26.6%4.8xPast-due book above 5.0%
ENGENCB 26, Engencap3.9%23.4%5.9xNot stated in the summary
HIRCB 25, HIR Casa4.2%29.3%7.0xAdjusted delinquency above 7.0%
TIPMXCB 23, TIP Auto1.0%22.1%22.5xBalance 120+ days late above 8.0%

None of these deals tests net loss over original principal. Their triggers read a stock of late loans, so the same history has to be restated before it answers a US-style schedule.

The triggers are the clearest difference. In HR Ratings' August 2025 review, TIP Auto's certificates name theirs «Pérdidas Acumuladas», accumulated losses, but define it as the balance of assets more than 120 days late over total assets, a stock read on each date; above 8.0% for more than two consecutive payment dates in the revolving period, amortisation accelerates. Exitus tests a cartera vencida ratio of 5.0%. Exitus's cohorts count a default at more than 90 days and leave out eleven clients who paid in full, on average 1.37 months after falling past due.

So the restatement a fund will ask for is mechanical. Take each monthly vintage, mark a loan defaulted at the day count the term sheet uses, subtract every recovery on it, and divide the running total by the vintage's original principal. Lay the curves on the proposed schedule month by month. The gap at month 6 and at month 12 is the headroom the lender is negotiating.

Definitions and trigger levels are the credit documents', the Mexican pools HR Ratings'; the worked pool, the timing curve and the shares of the schedules are our computation from them.

Questions this raises

What is cumulative net loss?
The principal of every loan in a pool that has defaulted since the cut-off date, less recoveries on those loans, divided by the pool's original principal. Credit documents test it pool by pool and month by month.
What is the formula for cumulative net loss?
CNL equals cumulative defaulted principal minus cumulative recoveries, divided by original pool principal. The NetCredit 2023 indenture counts a loan as defaulted at 65 or more days past due, nets liquidation proceeds and divides by the original principal sold to the issuer.
What is a cumulative net loss trigger?
A schedule of maximum CNL levels by month of the pool's life. NetCredit's 2023 indenture rises from 5.00% in the first collection period to 45.00% from the twenty-fourth; above it, an amortisation event sends all remaining collections to repay the notes.
Why divide by the original balance and not the current one?
The current balance falls as loans pay down, so losses over it rise even when the pool performs as expected. In a modelled two-year pool, 5.75% of original principal lost by month 12 reads 13.01% of the balance still outstanding, by our computation.
How do you project lifetime CNL from a young pool?
Divide its current CNL by the share of lifetime loss older pools had reached at the same month, the timing curve. The 2024 credit agreement between Consumer Portfolio Services and Citibank uses this projection in its advance rate calculator.
Do Mexican securitisations use CNL?
The rated ones HR Ratings reviewed in 2025 and 2026 compare a stress loss with a gross historical default rate by cohort, and their triggers read a stock of late loans, such as the balance over 120 days late at TIP Auto, 8.0%.
Where every figure on this page comes from 9 sources
  • 01NetCredit Combined Receivables 2023, LLC, Indenture dated 3 March 2023, Appendix A: Charged-Off Receivable, Cumulative Net Losses, Cumulative Net Loss Ratio, Cumulative Net Loss Trigger, Amortization Event. Enova International, 10-Q filed 24 July 2024, exhibit 4.3.
  • 02Conn's Receivables Funding 2023-A, LLC, Series 2023-A Supplement dated 17 August 2023: Cumulative Net Loss Percentage, Cumulative Net Loss Trigger, Annualized Net Loss Trigger. Conn's, Inc., 8-K filed 22 August 2023, exhibit 4.2.
  • 03Oportun CL Trust 2023-A, Amendment No. 1 to Receivables Loan and Security Agreement dated 22 March 2024: Amortization Period Cumulative Net Losses, CNL Trigger Test (levels redacted). Oportun Financial, 10-Q filed 10 May 2024, exhibit 10.5.
  • 04Page Eight Funding LLC, Consumer Portfolio Services and Citibank, Eighth Amended and Restated Credit Agreement dated 1 November 2024: Cumulative Net Loss Rate, Projected CNL, Projected CNL Matrix, Exhibit D timing curve (redacted). Consumer Portfolio Services, 10-K filed 16 March 2026, exhibit 10.1.8.
  • 05HR Ratings, CONCRCB 26S (FinCrementar), stock exchange notice of 19 August 2026.
  • 06HR Ratings, EXITUCB 26 (Exitus Capital, S.A.P.I. de C.V., SOFOM, E.N.R.), initial rating report of 26 May 2026: vintage analysis and trigger de cartera vencida.
  • 07HR Ratings, ENGENCB 26 and 26-2 (397 CAP, S.A. de C.V., SOFOM, E.N.R., and Engencap Fin), initial rating report of 27 April 2026.
  • 08HR Ratings, HIRCB 25 and 25-2 (Bienes Programados, HIR Casa), review of 24 June 2026.
  • 09HR Ratings, TIPMXCB 23 and 23-2 (TIP Auto, S.A. de C.V., SOFOM, E.N.R., and TIP de México), review of 22 August 2025: glossary (Pérdidas Acumuladas, Tasa de Incumplimiento Histórica) and trigger de pérdidas acumuladas.

Next step

If you lend

We rebuild your book from primary documents and bank movements, so a fund can read the evidence itself.

Create a lender accountHow Sylvent prepares a lender for a credit line

If you finance lenders

Read a lender's rebuilt book, monthly reports and flagged exceptions in a cabinet of your own. For a fund it is free.

Create a fund accountHow a fund monitors the lenders it finances
Capital. 12, Days a regulated SOFOM has to file its loans.

Where do a Mexican lender's numbers live at the CNBV, and how current are they?

The twelve- and twenty-day reports a regulated SOFOM owes the CNBV, what the Portafolio de Información republishes and how far behind, and the monthly grade the CNBV gives each lender's reporting.

Read it
Capital. 1,777, SOFOM E.N.R. with a loan book, March 2026.

What is a SOFOM, and how do a SOFOM E.N.R., a SOFOM E.R. and a SOFIPO differ?

A registration, a supervised regime and a deposit licence: the three routes that make a SOFOM regulated, what the CONDUSEF register shows for 364 unregulated ones, and what each form files in public.

Read it
Capital. 44.9%, Highest weighted gap, 4 Mexican lenders.

How do you calculate a lender's liquidity gap?

Assets maturing minus liabilities falling due, bucket by bucket: HR Ratings' two weighted ratios for four Mexican lenders in 2026, the ladders behind two of them, and the 30-day limit a regulator sets.

Read it
Capital. 3.3%, Average ROA, 84 Mexican lenders, 2026.

How do you calculate a non-bank lender's return on assets (ROA)?

Net income over average total assets: 3.3% for 84 Mexican non-bank lenders in 2026, why the same profit reads 4.1% over the loan book, where the return goes, and the covenant that caps it at 10%.

Read it
Capital. 48.7%, Efficiency ratio, 84 Mexican lenders, 2026.

How do you calculate a non-bank lender's efficiency ratio?

Operating expense over income before provisions: 48.7% for 84 Mexican non-bank lenders in 2026, why the same costs read 64.7% or 9.0% on other denominators, and how the segments change places.

Read it
Capital. 2.5%, One-year cost to hedge, Mexico, Oct 2026.

How do you calculate the cost of hedging a dollar credit line?

The forward premium is the gap between two interest rates: 2.5% a year for Mexico, 9.5% for Brazil and 0.6% for Chile in October 2026, what a hedged and an unhedged line cost, and what Mexican lenders carry.

Read it
Risk. 5.5%, PAR 30 of a book with 2.0% past 90 days.

How do you calculate portfolio at risk (PAR 30) on a loan book?

The whole balance of loans more than 30 days late over the gross portfolio: the two standard definitions, the 5% to 10% covenants funds write, and one Mexican book that reads 2.0%, 5.5% or 17.5%.

Read it
Capital. 13.0%, default ratio that stops a line revolving.

What covenants a fintech lender's first credit line carries, with the levels in filed agreements

Four kinds of covenant, read from three warehouse lines filed with the SEC: an advance of 50% to 80%, a first trigger at a 13.0% default ratio, net worth and cash tests on the parent.

Read it
Risk. 16.6%, defaults over 36 months at 10% CDR, 0% CPR.

How do you calculate CPR and CDR on a loan pool?

Monthly prepayments and defaults annualised by the SIFMA formulas, the CPR and default rates US consumer lenders file, Fitch's prepayment stresses for Latin America and what a CDR means over a pool's life.

Read it
Risk. 71%, floor reserve on a defaulted personal loan.

How do Mexican lenders calculate loan loss reserves?

Probability of default times loss given default times the balance, with the regulator's fixed loss tables for banks and regulated SOFOMs, an own expected-loss model for the rest, and the coverage the rated sector holds.

Read it
Capital. 100%, of each loan funded by the buyer at sale.

How does a forward flow agreement work for a lender?

An investor buys each new loan outright at a formula price, for a fixed term: the price, the share, the clock and the exits in Carvana's, Signet's and Pagaya's deals, and the Mexican trust that does the same job.

Read it
Capital. +6.25, points over TIIE on a Mexican SME line.

What does a first warehouse line cost a Latin American lender?

A base rate plus a margin on the money drawn, and fees on the rest: the margins in four filed credit lines, the fee schedule that makes an empty line expensive, and what one Mexican SME lender paid lender by lender.

Read it
Capital. 2.2x, leverage of 84 rated Mexican lenders.

How to calculate a non-bank lender's leverage ratio, and the levels Mexican lenders run at

Average liabilities less repo borrowing, over average equity: the formula a Mexican rating agency applies, reconciled to its balance sheet, the level by segment, and the capitalisation at which named lenders' ratings would move.

Read it
Capital. 3.75%, floor under a three-year pool's margin.

How to calculate excess spread on a loan pool, and the floor a credit line sets under it

Cash collected, minus every senior fee and the line's own interest, over the loans that still pay, times 12: the definition in two filed credit lines, the floor by loan term, and the margin Mexican non-bank books leave.

Read it
Risk. 72 months, before a bureau deletes a default.

What a Mexican lender reports to the credit bureau, and how it proves it every month

Buró de Crédito and Círculo de Crédito, the duty to report every loan with two dates, the bureau document filed with CONDUSEF monthly, and how long a default stays on file.

Read it
Risk. 57.9%, CAT of a 46.6% rate before any fee.

What the CAT (costo anual total) is, and what it leaves out of a Mexican lender's price

Banco de México's compounded, all-in annual cost of a loan: what Circular 21/2009 counts, why VAT, late fees and default stay out, and how a fund rebuilds the advertised CAT from the loan tape.

Read it
Capital. 1.30x, collateral behind each unit of debt.

What a fideicomiso is in Mexican lender financing, and the advance it carries

How a Mexican trust holds a lender's loans for the fund that financed them, and the aforo four deals with published terms required: 1.27x to 1.30x, an advance near 77% to 79%.

Read it
Risk. 90, days late moves an instalment loan.

What cartera vencida means for a Mexican lender

The CNBV day counts that move a Mexican loan into cartera vencida, why banks call it etapa 3 since 2022, and why an unregulated SOFOM's number needs its own rule checked.

Read it
Capital. 1.0x, when a cohort has paid back what it was lent.

How to measure collections from a loan tape

Cash a cohort has paid back, divided by the principal it was lent, read by months since origination. Why collection efficiency answers a different question, and the three tape mistakes a fund checks for.

Read it
Capital. 5 years, of cohorts the SEC and agencies ask for.

What a static pool is

A cohort of loans from one period, measured against its own original principal for life. Why the denominator is the point, what the SEC and the Mexican agencies ask for, and how an agency turns it into one rate.

Read it
Capital. 69 fields, in one template. 24 wins a term sheet.

What a loan tape is

One row per loan at a stated cut-off date. What the file has to carry, how the fields change by asset class, and why a tape without a data dictionary cannot be checked.

Read it
Capital. 28 items, before a first institutional facility.

The 28 items in a fund's first request

Eight lettered blocks and 28 numbered items, in the order a fund's own analyst files the answers, and 8 of them ask only that you reconcile your own documents against each other.

Read it
Risk. 4 documents, between one audit and the next.

How an investor watches a book it does not own

Four signed artefacts on a contractual cadence, recomputed rather than read, with consequences tied to arithmetic. Monitoring from the seat that has no access.

Read it
Capital. 37 questions, before anyone asks about yield.

What a fund asks on the first call, and the order of everything after it

The loan tape before yield, then six gates a credit committee clears one at a time, 37 items in all, and which of them a connected loan tape closes without anyone typing.

Read it
Risk. 11 months, between two verified numbers.

The eleven dark months

An audit reads a closed period once a year. What a fund prices is the gap between two audit dates, and that gap is where the failures this market remembers were found after the fact.

Read it
Mexico. 559 lenders, in one official count, not in the other.

Mexico cannot count its own lenders

Two official counts of the same sector, 559 apart, and no published loan book behind either. Who lends to Mexican small businesses, on whose money, and where the money stops.

Read it
Talk to us

Book a demo.

Tell us how you lend and what the capital is for. We will show you the platform on your own numbers and give you an honest read.

A founder reads it and answers within one business day, with a time that suits you.

  • ✓Thirty minutes, nothing prepared
  • ✓An honest read before you spend months on diligence
  • ✓We say no when the mandate does not fit

Still loading. The button switches on by itself and nothing you have typed is lost. If it never does, this browser has JavaScript turned off, and this form needs it.

One business day, and nothing to prepare.

By sending this form you agree to our Privacy Policy.