Same loans. Same market. Better outcomes.
Karus ran in parallel with a 15-person manual underwriting team for 20 months and $76.9 million of originations, on the same loan population, the same dealers, and the same conditions. Only the decision-maker changed.
20 bps higher
Seasoned-cohort average NUY (6.17% vs. 5.96%) with comparable cohort-level dispersion (stdev 0.35% vs. 0.34%)
.svg)
A controlled test, not a demo
The lender did not flip a switch. They ran Karus in parallel with their 15-person manual underwriting team for 20 months, on the same loan population, the same dealers, and the same market conditions. Every Karus decision was tracked against what the manual team did on the same deals. Only the decision-maker varied.
Both books originated at roughly 10.5% gross APR, within 2 basis points of each other. That parity is the control: with pricing held even, any difference in yield comes from selection quality, which loans each strategy chose to approve, not from charging more.
.svg)
.svg)
.svg)
.svg)
The result in five numbers
20 bps higher
Seasoned-cohort average NUY (6.17% vs. 5.96%) with comparable cohort-level dispersion (stdev 0.35% vs. 0.34%)
16 of 20
Monthly vintages where Karus delivered higher projected NUY
55% lower
Projected lifetime cumulative net loss (0.52% vs 1.17%, pooled)
66% lower
Unit default rate vs manual underwriting (0.43% vs 1.24%)
6× lower
Realized net charge-offs to date (0.07% vs 0.44%)
How gross APR became net unlevered yield
The static-pool waterfall shows where the yield difference came from. Origination, premium, and servicing are held equal across both strategies, which isolates the entire result to the loss line.
A 9 basis point advantage, the entire gap driven by the loss line.
The results, line by line
Pricing parity
Karus originated at 10.44% average APR versus 10.46% for the manual team, a 2 basis point gap. It reached that parity while underwriting borrowers with $1,354 lower average monthly income ($7,868 versus $9,222) at a comparable average credit score (708 versus 712). The pricing control holds, so the comparison stays clean.
Losses, by timing and severity
Across the full evaluation, Karus realized 0.07% in net charge-offs versus 0.44% for the manual team, a 6x difference in dollar terms, and a 0.43% unit default rate versus 1.24%. Early delinquency was roughly equal; but Karus loans rolled from delinquency to default at materially lower rates and produced smaller losses when defaults did occur.
Yield, vintage by vintage
Across seasoned cohorts, Karus delivers a ~20 bps higher average NUY with comparable cohort-level dispersion (stdev 0.35% vs. 0.34%) and a worst-case cohort (5.45%) at roughly the level of manual underwriting's average
We ran the downside too
The Karus book is younger than the manual book, so we tested the conclusion against its own weakest case. In the stress scenario, Karus lifetime losses are forced to converge to the manual book's projected lifetime cumulative net loss of 1.17%, as if Karus selection were no better than manual at all.
| Karus Base | Karus Stress | Manual | |
|---|---|---|---|
| Average APR (dollar-weighted) | 10.44% | 10.44% | 10.46% |
| Pooled lifetime CNL | 0.52% | 1.17% | 1.17% |
| Projected net unlevered yield | 5.95% | 5.83% | 5.86% |
Where manual won
Manual underwriting delivered higher projected yield in 4 of the 20 vintages. In each case the gap came from APR pricing rather than loss performance: Karus priced those cohorts lower, often by 30 to 50 basis points, and the youngest 2026 vintages are too unseasoned for realized losses to mean much yet. We show them rather than exclude them.
Even at loss parity, Karus projected net unlevered yield is 5.83% against the manual team's 5.86%, within 3 basis points. The upside is real; the downside is approximate parity, not underperformance.
Why predictable yield matters as much as yield
01
Predictability Pays
In structured finance, volatility is a hidden cost. Wide swings in losses force larger reserve buffers for the downside. Karus is projecting 0.52% pooled lifetime cumulative net loss versus 1.17%, with a tighter loss distribution across vintages.
02
Consistency Earns Better Terms
That predictability does two things. It lets a lender reduce reserve buffers and redeploy that capital into new originations. And it lets warehouse lenders and securitization investors price the paper as a programmed outcome rather than a range. Certainty earns better terms.
“We came to Karus to standardize underwriting. What we found changes how we manage portfolio performance: consistent outcomes, predictable yield, and the ability to grow without losing control.”
Auto finance originator, 20-month Karus evaluation
See the same analysis on your portfolio
We will rebuild your book loan by loan through Karus and show you the projected net unlevered yield, the loss curves by vintage, and where the selection differs from your current process.