Reading consumer lender loss rates through unit economics

The board pack is open the night before the risk committee. A peer comparison pasted from listed lenders’ results puts your loss rate beside a lender that loses a large share of its book each year and another that loses very little. A board member has circled the line. In the margin, the question is why yours is not the low one.

What does that line let you explain about the risk you chose to take?

Compare loss rates with each lender’s revenue

The low loss rate looks like the benchmark in a peer table. But the annualized net charge-off rate measures losses against average receivables and says nothing by itself about the revenue available to pay for those losses.

In the third quarter of 2026, OppFi reported a 52.6% annualized charge-off rate, Synchrony 4.79% and OneMain 6.89%. The comparison changes when each lender’s credit cost is set against its own revenue. Per 100 of revenue, OppFi booked 37.8 for risk, Synchrony 21.5 and OneMain 36.1. Each also reported a profit.

OppFi’s charge-off rate was many times Synchrony’s, yet the share of revenue each booked for risk differed by less. OneMain’s risk share was close to OppFi’s.

The figure compares how five lenders split 100% of revenue. OppFi, Q3 2026: risk 37.8%, funding 5.8%, acquisition 8.7%, running costs 22.8%, profit 24.8%. Its annual loss measure is 52.6%. Enova, Q3 2026: risk 39.1%, funding 10.3%, acquisition 22.8%, running costs 13.7%, profit 14.2%. Its quarterly loss measure is 8.1%. OneMain, Q3 2026: risk 36.1%, funding 20.3%, running costs, including acquisition, 29.9%, profit 13.7%. Annual loss measure: 6.89%. Synchrony, Q3 2026: risk 21.5%, funding 16.6%, acquisition 20.9%, running costs 20.9%, profit 20%. Annual loss measure: 4.79%. Qifu, Q2 2026: risk 29.9%, funding 3.6%, acquisition 11.1%, running costs 22.8%, profit 32.6%. Its loss measure is 2.83% over 90 days past due.
Revenue split per 100 for five lenders
lenderquarterloss on the bookriskfundingacquisitionrunningprofit
OppFiQ3 202652.6% a year37.85.88.722.824.8
EnovaQ3 20268.1% in the quarter39.110.322.813.714.2
OneMainQ3 20266.89% a year36.120.3in running29.913.7
SynchronyQ3 20264.79% a year21.516.620.920.920.0
QifuQ2 20262.83% over 90 days past due29.93.611.122.832.6

Read what each lender’s revenue can absorb

A lender carries risk from the revenue left after credit losses, funding, acquisition and running costs. A higher borrower yield can support a higher credit cost. The bad rate a lender can carry is a unit-economics decision as well as a credit decision.

The risk line is the quarter’s credit charge, which can include a reserve build and differ from charge-offs. Enova reported quarterly charge-offs, while Qifu’s figure is a 90 day+ balance rate. Synchrony’s acquisition grouping includes payments to retail partners.

When you assess your own book, compare price and loss together, as in risk-based pricing. The relevant bad rate is the one your yield and cost structure can support while still meeting your profit floor.

Set your loss tolerance from vintage losses

Set appetite against cumulative losses from your own lending periods, then check whether the revenue from those loans still meets the profit floor you accept.

A balance-based delinquency rate uses the outstanding balance on a date, so the rate can rise when that balance shrinks even if overdue principal stays the same. Use vintage analysis to read performance by origination period beside balance-based measures.

SoFi reported 4.71% net cumulative losses on recent personal-loan vintages originated from the fourth quarter of 2022 through the fourth quarter of 2025. It said those results supported a maximum cumulative net loss assumption of 7–8%. That leaves 2.29 to 3.29 percentage points of headroom to the stated maximum.

Put these checks into a decision flow

Create rules for credit, fraud or collections, then run them on live decisions.

Use market trends to decide what to check

OneMain’s annualized US charge-offs reached a high in the first quarter of 2024 and were lower by the third quarter of 2026. Synchrony’s rose through 2024 and then eased to 4.79% in the third quarter of 2026. SoFi’s personal-loan rate was lower than a year earlier, while OppFi’s was higher.

Line chart of net charge-off rates for OneMain and Synchrony. OneMain was 5.57% in 2022 Q1, reached a high of 8.58% in 2024 Q1, and was 6.89% in 2026 Q3. Synchrony was 2.18% in 2021 Q3, reached a high of 6.45% in 2024 Q4, and was 4.79% in 2026 Q3.
OneMain and Synchrony annualized net charge off rates

In China, Qifu, Lexin and FinVolution’s 90 day+ rates rose in early 2024, eased through much of 2025 and rose again in early 2026. Qifu and FinVolution’s rates were lower in the second quarter of 2026 than in the first quarter, while Lexin’s was higher. The measures have different definitions, so follow each platform against its own series.

90-day-plus delinquency rates. Qifu was 1.17% in 2021 Q3, peaked at 3.5% in 2026 Q1, and was 2.83% in 2026 Q2. Lexin was 1.85% in 2021 Q3, peaked at 3.7% in 2024 Q2, and was 3.6% in 2026 Q2. FinVolution was 1.04% in 2021 Q3, peaked at 3.11% in 2026 Q1, and was 2.1% in 2026 Q2.
90 day or more delinquency rates at three Chinese lenders

Qifu’s outstanding loan balance fell 23.2% between June 2025 and June 2026. Its 90 day+ rate divides overdue principal by outstanding principal at a date, so a shrinking balance can raise the ratio.

FinVolution reported month-twelve vintage delinquency of 2.36% for 2022 loans, 2.84% for 2023 loans and 2.32% for 2024 loans. Lexin’s chart shows its 2023 to 2024 vintages ending above its 2022 vintages. SoFi’s recent-vintage cumulative losses remained below its stated maximum assumption.

The chart compares FinVolution China loan vintages at month 12. The 2022 vintage is 2.36%, the 2023 vintage is 2.84%, and the 2024 vintage is 2.32%.
FinVolution delinquency by loan vintage at month twelve

Qifu reported that its day-one delinquency rose to 6.1% in the fourth quarter of 2025, while its 30-day collection rate fell to 84.1%. Those measures were 5.6% and 88.1% in the second quarter of 2026. A rise in overdue principal paired with weaker collection is a reason to review the affected vintages and update expected losses.

In Brazil, Nu’s risk-adjusted net interest margin was 12.1%, up from 10.8% a year earlier and down from 12.4% in the prior quarter. In Kazakhstan, Kaspi reported average funding cost of 14.9%, up 150 basis points year over year.

marketlendermeasurea year beforelatest
USSynchronynet charge-offs a year5.16%4.79% (Q3 2026)
USOneMainnet charge-offs a year6.67%6.89% (Q3 2026)
USSoFi personal loanscharge-offs a year2.60%2.26% (Q3 2026)
USOppFinet charge-offs a year46.7%52.6% (Q3 2026)
ChinaQifu90 day+1.97%2.83% (Q2 2026)
ChinaLexin90 day+3.1%3.6% (Q2 2026)
ChinaFinVolution90 day+1.92%2.10% (Q2 2026)
BrazilNuNPL 90+6.7%7.3% (Q3 2026)
KazakhstanKaspi.kzNPL ratio6.1% (2025)6.9% (9M 2026)

Recalculate your risk budget as costs shift

When price falls or funding costs rise, solve again for the risk cost your book can carry before changing the cut-off. Qifu’s second-quarter 2026 accounts show the arithmetic. Per 100 of revenue, its risk cost rose from 25.3 to 29.9 and operating profit fell from 41.0 to 32.6 year over year.

In Qifu’s second quarter of 2026, funding, acquisition and running costs together were 37.5 per 100 of revenue, versus 33.7 in the prior-year quarter. At any unchanged profit floor, the amount available for risk cost had therefore fallen by 3.8 per 100 of revenue.

Qifu’s annual report said its platform adhered to a 24% annualized IRR ceiling for new loans. Its later results described industry-wide funding constraints and tighter risk standards. Recalculate the risk budget using current revenue and costs, then test it against the losses of the vintages you would approve.

Set a repeatable loss rule for your book

Return to the board pack with a repeatable calculation using your own revenue, costs and vintage performance to show where the profit floor puts the limit and what would cause a review.

  1. Put your loss rate beside what you charge. Compare losses with yield or risk cost per 100 of revenue before comparing lenders.
  2. Split 100 of your revenue into risk, funding, acquisition, running cost and profit. Keep reported provisions or fair-value changes distinct from charge-offs.
  3. Subtract all costs and the profit floor from revenue per 100. Use vintage cash flows and losses to set cumulative tolerance.
  4. Record realized and expected remaining losses against each vintage’s initial principal. When volume changes, compare vintages with balance-based delinquency rates.
  5. Track day-one delinquency and 30-day overdue-principal collections monthly. If either worsens, review affected vintages and update expected losses before changing the cut-off.
  6. Recalculate per-100 revenue and risk budget when price caps or funding costs rise, then test loss tolerance against the floor before changing the cut-off.

Questions about consumer lender loss rates

Why do consumer lenders have such different bad rates?

Borrower yield and costs differ. Revenue covers risk, funding, acquisition and running costs, with profit left over. Compare risk cost with each lender’s revenue split before judging its loss rate.

What is a vintage delinquency rate and how is it different from a 90 day+ rate?

A vintage rate tracks loans from an origination period against initial principal. Qifu’s 90 day+ rate divides overdue principal by the outstanding balance on a date. Check each lender’s definitions.

Can a lender with a 50% annual charge-off rate be profitable?

Yes. OppFi reported a 52.6% annualized charge-off rate and operating profit of 24.8 per 100 of revenue in the third quarter of 2026.

How do you set the bad rate your lending can afford?

Calculate the risk-cost budget left after funding, acquisition, running costs and your accepted profit floor, then assess cumulative losses by vintage against that budget. Recalculate when price or funding costs change.

Put these checks into a decision flow

Create rules for credit, fraud or collections, then run them on live decisions.