The average hides the arsonist
Your book-level charge-off rate can look perfectly healthy while one dealer runs four times your average inside it. Averages don’t name names — and the dealer sending you poison knows it.
Dealer forensics for auto lenders
Which dealersare poisoning your book.
Dealer X-Ray traces every origination source to how its paper actually performed — headlined by the early-payment-default radar, the smoking gun for a toxic dealer or fraud. Transparent arithmetic on your own book, over a read-only join of your LOS and LMS.
An EPD — a loan that dies inside three payments — is almost never a credit miss. It's a deal that was wrong the day it was signed.
The first app lenders buy on the Underbot platform.
Traceevery number is arithmetic you can check
Read-onlyyour LOS and LMS, untouched
60 dayspilot start to keep-or-walk readout
Dealer scorecardSynthetic book · worst first
Example onlyBook EPD Rate2.1%
Toxic Dealers3
61+ DPD Now41
DealerTierEPD Rate×BookNet Loss
Dealer #1120Toxic9.8%4.6×$212K
Dealer #0847Watch4.4%2.1×$96K
Dealer #0533Standard2.1%1.0×$84K
Dealer #0291Star0.8%0.4×$6K
Dealer #1201Low Vol———
The smoking gunDealer #1120 runs 4.6× your book — its charge-offs average 1.8 payments before dying.
4.6×
In production with Tracir Financial Services, a multi-state auto lender
Live, read-only integration into your LOS and LMS — not batch extracts
Onboarding a limited number of lenders at a time
The blind spot
Your LOS knows who sent the deal. Your LMS knows how it died. Neither will tell you which dealers to cut.
The number that ends a dealer conversation lives in the join of the two systems — and until that join exists, every bad source hides in plain sight.
Fraud reads as bad luck, months late
A misrepresented deal surfaces in collections long after funding, with no line back to the desk that sent it. By the time the pattern is obvious, the next twenty deals from that desk are already booked.
Dealer terms get set on volume
Reps negotiate on units sent, relationships, and last month’s numbers. Nobody at the table has the figure that matters: how the last hundred deals from that dealer actually performed.
The EPD radar
A loan that dies inside three payments was never really a loan.
The radar's headline signal is the early payment default — three conditions, no judgment calls, straight off your own servicing records.
Boarded 6+ months ago+≤ 3 successful payments+Charged off, or 61+ days past due now=EPD
The same distress signal inside six payments is a Soft EPD — the early-warning band, so a dealer drifting toxic shows up before the hard number does.
Seasoning — 6 months before a loan counts
Young paper hasn’t had time to fail its first payments. Counting it would flatter exactly the dealers ramping volume right now — so every rate uses only loans boarded six-plus months ago.
Minimum sample — 10 seasoned loans to earn a grade
One bad loan out of three would scream 33%. Dealers below ten seasoned loans get their raw numbers but no tier — the math never slanders a small sample.
The grades
Every dealer graded against your own book. Not an industry table — your book.
Each dealer's EPD, charge-off, and loss rates become multiples of your book average, and the worst of the three sets the grade — a dealer can't hide a toxic EPD rate behind a decent loss number.
≥ 2.5× bookToxicCut, reprice hard, or claw back
≥ 1.5× bookWatchTier the relationship and watch it weekly
In betweenStandardBusiness as usual
≤ 0.6× bookStarSend them more volume
< 10 seasoned loansLow VolumeNumbers, no grade — too few to judge
The thresholds ship with these defaults and stay tunable — once you see real output on your book, you set where Watch ends and Toxic begins.
6 moseasoning before a loan counts
≤3payments defines an EPD
10seasoned loans to earn a grade
2.5×book average = Toxic
0black boxes — arithmetic only
The scorecard
One page. Every dealer. No place to hide.
This is what the pilot stands up on your book — filter by date range, search by dealer, click any row to open the distressed paper behind it. Synthetic numbers below; yours will have names on them.
Book EPD Rate2.1%
Charge-Off Rate6.4%
Net Loss$1.9M
Toxic Dealers3
61+ DPD Now41
| Dealer | Tier | Booked | Seasoned | EPD | EPD Rate | ×Book | Net Loss | 61+ DPD | Avg Pmts→CO |
|---|---|---|---|---|---|---|---|---|---|
| Dealer #1120 | Toxic | 212 | 164 | 16 | 9.8% | 4.6× | $212K | 9 | 1.8 |
| Dealer #0847 | Watch | 149 | 114 | 5 | 4.4% | 2.1× | $96K | 4 | 3.9 |
| Dealer #0533 | Standard | 301 | 240 | 5 | 2.1% | 1.0× | $84K | 5 | 6.2 |
| Dealer #0678 | Standard | 96 | 71 | 1 | 1.4% | 0.7× | $18K | 1 | 8.4 |
| Dealer #0291 | Star | 163 | 128 | 1 | 0.8% | 0.4× | $6K | 1 | 11.5 |
| Dealer #1201 | Low Volume | 11 | 7 | 1 | 14.3% | — | $9K | 0 | — |
Six of the book's 38 dealers shown — a slice across the grades. In the live scorecard the worst sort to the top automatically. Dealer #1201 stays ungraded: seven seasoned loans is too few to judge, even at a 14.3% raw rate.
61+ DPD Now
Tomorrow’s charge-offs, today.
Avg Pmts to CO
Low single digits means the paper was dead on arrival.
Emerging Risk
Also tracked per dealer: loans too young to count in the rates, but already dying.
Inside Dealer #1120 — the distressed paper5 of 24 distressed loans · fewest payments first
A-4471Nov 20250 pmtsEPDCharged Off$18,940
A-4508Oct 20251 pmtsEPDCharged Off$16,220
A-4610Dec 20251 pmtsEPDActive · 74 DPD$14,660
A-4655Sep 20252 pmtsEPDCharged Off$15,480
A-4702Nov 20255 pmtsSoftActive · 89 DPD$13,750
The closer to zero, the louder the smoking gun. This is the page you put in front of a dealer who wants to argue with his grade.
What lands on your desk
Not a dashboard subscription. Artifacts your team acts on.
Every dealer tiered and sortable, with date-range filters, dealer search, and the distressed-loan drill-down one click deep.
An Excel export — Dealer Scorecard and Distressed Loans sheets, with the methodology footnoted on every column. The artifact you put on the table when you call the dealer in.
Findings in plain lender English: which sources to tier, reprice, claw back, or cut — and what changed since last week.
The keep-or-walk number: what the radar found, what it’s worth on your book, and whether the platform earns year one.
No black box
Arithmetic you can check. Not a model you have to trust.
Dealer X-Ray is ratios and multiples on your own loans — no machine learning, no probability score, nothing trained on your data. Payment counts come from your servicing transactions, outcomes from your servicing system, financed amounts from your loan terms — joined read-only to the origination record that names the source. Any analyst with the workbook can recompute every rate.
The radar flags a dealer for investigation. Your team makes the dealer call.
Read-only, always
Your book is the only benchmark
6-month seasoning before a loan counts
10-loan minimum before a grade
Thresholds you can tune
A workbook that shows the math
No machine learning, no probability score
Nothing trained on your data
Who reaches for it
One scorecard, four people who stop guessing.
One page that names the sources behind the losses — and puts a defensible number on every dealer relationship the company keeps.
Policy says one thing; funded paper says another. The radar shows which desks send deals that die on contact, in your own book’s arithmetic.
The workbook is the conversation: boarded dates, payment counts, exposure, fewest payments first. The grade stops being your opinion.
Net loss by source — unrecovered deficiency balances, net of recoveries to date — instead of a gut feel about which relationships pay for themselves.
The 60-day pilot
Sixty days on your full book. A keep-or-walk number at the end.
Dealer X-Ray is rung two of how lenders buy Underbot — after the free teardown, before the platform. The pilot is scoped to end with a decision, not a renewal conversation.
The free teardown — ten anonymized files on our template, under NDA, plus a findings memo and a 30-minute walkthrough. No integration, no committee, no invoice.
We stand up the read-only LOS + LMS join and run your full book through the EPD radar — every dealer tiered against your own averages, with the workbook and weekly memos along the way.
Loss Radar, the Collections War Room, the Underwriting Engine — each added on the join your pilot already built and proved on your paper.
What the pilot stands up
- Your live, read-only LOS + LMS join, stood up on your own systems — the same plumbing the whole platform runs on
- Your full book through the EPD radar — every dealer tiered Toxic, Watch, Standard, or Star against your own averages
- The distressed-loan drill-down behind every grade — boarded dates, payment counts, status, exposure
- The Excel workbook your team can put in front of a dealer, methodology footnotes included
- A weekly findings memo in plain lender English — tier, reprice, claw back, or cut
- A day-60 readout with the number either way: keep going or walk away
The terms
60 days, start to readoutOne fixed fee — integration includedCredited 100% toward your first yearRead-only, on feeds you already produce
The fee is quoted at your free teardown readout — one fixed number, integration included, credited 100% toward year one if you continue.
Start with the free teardownStraight answers
The questions a careful lender actually asks.
Is this a model telling us to fire dealers?
No — it’s arithmetic. Every number is a ratio your analyst can recompute from the workbook: EPDs over seasoned loans, multiples against your own book average. The radar flags a dealer for investigation — your team makes the dealer call.
Our dealers will fight the numbers.
Good. Bring the drill-down. Behind every grade sits the loan-level record — boarded dates, payment counts, status, exposure, sorted fewest payments first. A dealer can argue with a grade. It’s much harder to argue with sixteen loans that died inside three payments.
Can’t our LMS already report this?
It doesn’t know who sent the deal. Your servicing system sees payments but not sources; your origination system sees sources but not payments. The radar runs on the read-only join of the two — that’s the whole trick, and it’s why this number doesn’t exist in either system alone.
Most of our dealers are small.
The math never slanders a small sample. Dealers with fewer than ten seasoned loans get their raw numbers but no grade — one bad loan out of three would otherwise scream 33%. Grades start once a dealer has real volume to be graded on.
What do you need from us to start?
Read-only access to feeds you already produce. We stand up the LOS + LMS join on your own systems — no writes, no workflow changes, no new exports for your team to build. Sixty days later you have the readout.
What happens at day 60?
You get the number either way. The readout names what the radar found and what it’s worth on your book. If the case isn’t there, you walk away and keep the findings. If it is, the pilot fee is credited 100% toward your first year.
Is this the same thing as Loss Radar?
No — same join, different job. Dealer X-Ray looks backward: which origination sources produced the paper that died. Loss Radar looks forward: which live loans are drifting toward charge-off right now. Most lenders start here, because the dealer decision stops the bleeding at the source.
Start where every lender starts
Score your last 10 deals. Free.
The ladder starts with the free teardown — ten anonymized files, a findings memo, and a 30-minute walkthrough of what your own book has been trying to tell you. If the case is there, the Dealer X-Ray pilot is the next rung.
Free teardown firstRead-only when you connectFee credited 100% toward year oneKeep-or-walk number at day 60