Risk rating validation every night: where like borrowers get different grades, and why
Grade drift seen within the quarter, before the examiners pull their sample.
See one case, screen by screen ↓Software that checks whether the risk grades on your loans are accurate and consistent. Risk Rating Validation re-scores every rated commercial loan on your own scorecard each night, compares it with look-alike borrowers graded by other lenders and in other regions, and reads the memo behind each grade — so you see where like borrowers are graded apart, and why.
How to choose AI solutions for lending →
Updated
Every commercial loan carries a grade, and the grade drives the allowance, pricing, approval authority and how often the credit is looked at. The grade is set by people — often the lender who brought the deal in. Two lenders in two regions can grade near-identical borrowers two notches apart, and the memo rarely says why: it leans on “a long-standing relationship” and a strong sponsor instead of the ratios the scorecard used, and the override goes in with no reason code.
Loan review sees a sample — at Kestrel Valley Bank, 401 of 2,305 rated loans in a quarter. Checking like against like across regions means re-spreading and re-scoring borrowers side by side, so analysts do it for a handful of loans, never the whole book. Drift in one region and one industry can run for a year before a sample lands on it, and it often shows up as a double downgrade. When examiners find more than 5% of the loans they sample graded wrong, or 3% of the dollars, they look for the root cause and may widen the sample, and the rating system itself can become a finding.
Estimated split, in days, for one quarterly calibration of a single drifting region — by hand on a handful of loans, and with the solution on every loan.
Six specialist agents re-score every loan and pair look-alike borrowers graded apart; a regional credit officer and the CCO decide.
Freight trucking in the Mountain region is graded better than the same borrowers anywhere else in the bank. Here is how Rachel Okafor finds it, proves it and takes it to Senior Credit Committee, screen by screen, in the working solution.
Rachel opens the home screen: 195 rated freight trucking loans in five regions, each borrower placed by its debt service coverage and senior debt to EBITDA, coloured by the grade it was given, over shaded bands for the grade the bank’s scorecard gives. Two dots sit almost on top of each other and are circled: Ridgeback Freight Lines, graded 4 by Brett Kalani in Denver, and Coulter Haulage, graded 7 by Linda Ferraro in Omaha.
“Coverage within 0.03x, leverage within 0.1x, earnings down 18% and 15%. The scorecard puts both at 6. Ridgeback's 4 is an override with no reason code; Coulter's 7 is a documented covenant downgrade.”
She recolours the dots by notches off the scorecard and picks the Mountain region. Fourteen loans are graded two or more notches better than the bank’s own scorecard — and 11 of them are with one lender, Brett Kalani. None of its 38 trucking loans is graded stricter than the scorecard.
Ridgeback (loan C-25-04417, $18.5M: $12.0M term and a $6.5M revolver) sits beside its matched peers — Coulter in Omaha, Thorne Freight Systems in Milwaukee, Brazos Line Carriers in Waco. Revenue, coverage and leverage all match within tolerance. Assigned 4; scorecard v4.2 says 6; peers 7, 6 and 5, median 6.
“Graded 4 here, 7 in Omaha — for the same numbers.”
The scorecard shows factor by factor how Ridgeback scores 5.6, a grade 6, and the lender’s override of −2 notches with reason code “none”. The annual review affirms grade 4 on “a long-standing relationship since 2014” and a sponsor with “substantial outside liquidity”. It scores 34 of 100. Coulter’s covenant review, which explains its downgrade, scores 82.
Missed: DSCR below the 1.25x guide not explained · EBITDA −18% not mentioned · guarantor statement from Jun 2024 · override without a reason.
The drift map lays nine industries against five regions — 2,305 loans in 45 cells. Four cells lean more than half a notch from their peers. Freight trucking in the Mountain region is the widest: 38 loans, $286.4M, 1.2 notches better. Brett Kalani’s loans sit 1.6 better, Sara Lindqvist’s 1.0, Owen Pratt’s 0.7.
“Grades lean on relationship and sponsor support; 19 overrides of the scorecard carry no reason code. Loan review downgraded 3 of these loans in Q3.”
Rachel runs the comparison on the 38 loans. The agents read 38 grades and 61 grade changes since 2023, put the borrowers on one basis, apply scorecard v4.2, find 142 like-for-like borrowers in the other four regions and score 38 memos. Each step shows what it found: 30 of 38 grades set by the lender alone, 6 guarantor statements older than 15 months, “long-standing relationship” in 17 memos.
“14 loans, $146.8M, two notches better · 11 are Brett Kalani's · propose 12 regrades, keep 2.”
Loan review changed 5 of 82 sampled Mountain grades in Q3 — 6.1% of loans and 4.5% of dollars, above the 5% and 3% lines examiners watch. All five were downgrades, three of them trucking. The region upgraded 1.8 loans for every downgrade while the bank ran at 0.9, and its grade 4 loans defaulted at 1.9% over three years against 0.8% expected.
The drafter writes the finding and its cause, every line cited: Mountain loans under $20M are graded by the lender alone, 19 of 22 overrides carry no reason code, and spot-rate exposure is not a factor in scorecard v4.2. Of the 14 loans it proposes 12 regrades — 7 to grade 6, 2 to grade 5, 3 into Special Mention ($31.4M) — and keeps Cache Valley Freight and Snake River Carriers, whose memos show why.
Four policy changes, starting with “A credit officer co-signs every override of the scorecard”. Allowance effect +$2.9M — an estimate, for the allowance committee to measure.
Rachel sends the note for sign-off. Kevin Brandt confirms the 12 regrades — he can keep or change any grade, and the lender on the loan never confirms his own. Nothing changes in the loan system until committee approves.
Daniel Whitford signs the Q3 calibration note for Senior Credit Committee. The 12 regrades are queued for the loan system the morning after committee, Brett Kalani and Sara Lindqvist are notified, and the drift map re-runs that night. Every step is on the activity trail.
“On the agenda for Tue 20 Oct. 12 regrades queued for the loan system the morning after committee; Brett Kalani and Sara Lindqvist notified; the drift map re-runs tonight.”
The same drifting region, seen by the four people who carry it — what their quarter looked like, and what it looks like now.
Six specialist agents re-score every loan and pair look-alike borrowers graded apart; a regional credit officer and the CCO decide.
Reads every grade and grade change from the loan system — who set it, when, the scorecard grade at the time, the override and its reason code — and flags overrides without one and grades set by the lender alone.
Puts each borrower’s latest statements on one basis — DSCR, senior funded debt to EBITDA, fixed charge coverage, current ratio, liabilities to tangible net worth, EBITDA trend — and dates the guarantor statement.
Applies the bank’s approved scorecard (v4.2) factor by factor with its weights and grade guides, and reports the weighted score, the scorecard grade and the gap to the grade given.
Finds 3 to 5 like-for-like borrowers for each loan — same NAICS 3-digit industry, revenue within 25%, DSCR within 0.10x, leverage within 0.5x — graded by a different lender, and reports their grades and median.
Reads the approval memo or annual review that set or affirmed the grade and scores it on five tests: ratios cited, each weakness addressed, override explained with a reason code, current statements used, and what would move the grade.
Writes the finding and its cause, a proposed grade for each flagged loan, policy changes and the estimated effect, and routes the note to the regional credit officer, then the Chief Credit Officer.
Confirm each regrade; sign for committee; never the lender. The agents propose; a named person decides.
Rachel and the credit officers can ask in plain words — or ask for the note to be drafted. Every answer points to its source.
Why is Ridgeback Freight graded better than its peers?
Ridgeback carries grade 4. The bank’s scorecard gives it 6 (weighted 5.6), and its three closest trucking peers — Coulter in Omaha, Thorne in Milwaukee, Brazos in Waco — are graded 7, 6 and 5. The gap is a two-notch override by Brett Kalani with no reason code. The memo rests on “a long-standing relationship” and a guarantor statement from June 2024, so it scores 34 of 100.
Show the Mountain trucking loans two notches better than their peers
14 loans, $146.8M: 11 with Brett Kalani, 2 with Sara Lindqvist, 1 with Owen Pratt. Three of them — Pinecrest, Arapahoe and Big Horn — have coverage under 1.05x and sit with peers graded Special Mention.
Would Mountain’s grades pass the examiners’ sample?
Not as they stand. Loan review changed 5 of 82 sampled Mountain grades in Q3 — 6.1% of loans and 4.5% of dollars. Examiners treat more than 5% of loans or 3% of dollars as significant and look for the root cause. The bank as a whole is at 3.7% and 2.4%. The exam starts 8 Feb 2027.
Draft the calibration note for Tuesday’s committee
Drafted for Senior Credit Committee, Tue 20 Oct. Of the 14 loans: 12 regrades — 7 to grade 6, 2 to grade 5 and 3 to Special Mention — and 2 kept, because Cache Valley’s and Snake River’s memos show a verified guarantee and a contract. Four policy changes, starting with a credit officer co-signing every override. Kevin Brandt confirms the regrades; Daniel Whitford signs.
12 screens from the working solution, on its sample data. Pick one to see it large.
How borrowers are matched, when a grade is flagged and who decides are settings — changes are versioned and on the audit trail.
| Setting | Default | Choose from |
|---|---|---|
| Debt service coverage tolerance | ±0.10x | ±0.05x · ±0.10x · ±0.15x · ±0.20x |
| Leverage tolerance (senior funded debt to EBITDA) | ±0.5x | ±0.25x · ±0.5x · ±0.75x |
| Size band | Revenue within 25% | Within 15% · 25% · 40% |
| Flag a cell that leans from its peers by | 0.5 notch | 0.3 · 0.5 · 0.75 notch |
| Flag a single loan graded from its peers by | 2 notches | 1 · 2 · 3 notches |
| Memo score minimum | 60 of 100 | 50 · 60 · 70 of 100 |
| Largest override without a credit officer | 1 notch | None — credit officer co-signs every override · 1 notch · 2 notches |
| Run | Nightly and on every new grade | Nightly and on every new grade · Weekly · Month end |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample bank data · “target” = the design goal, measured in the live solution · “estimated” = our estimate · “typical” = published figures (OCC Comptroller’s Handbook, Rating Credit Risk — above 5% of credits or 3% of dollars graded wrong, examiners look for the root cause and decide whether to widen the sample; 12 U.S.C. 1820(d) — full-scope exams every 12 months, 18 for qualifying smaller banks). People, banks and borrowers named on this page are fictional — characters and sample data in the working solution.
Software that checks whether the risk grades on your loans are accurate and consistent. Risk Rating Validation re-scores every rated commercial loan on your own scorecard each night, compares it with look-alike borrowers graded by other lenders and in other regions, and reads the memo behind each grade — so you see where like borrowers are graded apart, and why.
The peer matcher pairs each borrower with like-for-like borrowers graded by a different lender: same NAICS 3-digit industry, revenue within 25%, debt service coverage within 0.10x and leverage within 0.5x. Where the grades sit two notches or more apart, the pair is shown side by side. All tolerances are settings.
No. The agents read the loan system and never write to it. They propose regrades in a calibration note; nothing changes in the loan system until committee approves.
A regional credit officer confirms each regrade and can keep or change any of them — never the lender on the loan. The Chief Credit Officer signs the calibration note for Senior Credit Committee. People decide every grade.
Yes. The scorecard re-scorer applies your approved version, factor by factor with its weights and grade guides, and never changes them. The version in use is a setting: here, v4.2 approved in January 2026, with a v4.3 draft that adds spot-rate exposure for truckload carriers going to model validation before use.
On five tests worth 100 points: it cites the ratios behind the grade, addresses each weakness the scorecard found, explains any override with a reason code, uses current statements including the guarantor’s, and says what would move the grade. It never judges the credit itself, and quotes the passage for each finding.
It watches loan review disagreement by region against the examiners’ lines — 5% of loans or 3% of dollars — alongside upgrades per downgrade, double downgrades and a three-year default back-test by grade, so drift is seen within the quarter rather than at the next exam.
The Agentic Solution Engine builds and deploys it from your requirements — your rating scale, scorecard, credit policy and a sample of rated loans and memos — and it goes live once every quality gate has passed. We will walk you through it on your own book first.
We’ll run Risk Rating Validation on a sample of your own rated loans, scorecard and memos.