# Auto Credit Intelligence · System the deal population, bands and validation. Live page: https://getbookiq.com/intel/#system --- What this is. The statement-view sibling of the credit union tape room, applied to the SEC filers that hold a consumer credit book. NCUA hands every credit union one account dictionary. The SEC does not, so the dictionary here is a versioned mapping of normalized fields to the concepts each filer actually tags, and the mapping is the product. Where the CU build paraphrases definitions, this one carries the taxonomy's own label and description for whichever concept resolved, so the bottom of every statement read shows you exactly what you are looking at and under which tag. Choosing the concept. A field resolves to one concept and one only, ranked by coverage in the recent window, then lifetime coverage, then preference order. Two failures forced that rule. Filling gaps from a second concept spliced Conn's total allowance onto a current-portion-only tag and manufactured a collapse from 35% coverage to 0.08% and back, which is indistinguishable from a credit event. Ranking on lifetime coverage alone picked concepts World Acceptance and Credit Acceptance abandoned years ago and left the current tape empty. Provenance, and why item codes are not scored. Outcome events are hand-curated and every one was confirmed by fetching and reading the filed document. EDGAR's submissions metadata is not reliable for this: Regional Management's 2024-06-20 filing is tagged with Item 1.03, Bankruptcy or Receivership, but the document is a ninth amendment to a revolving credit facility. Scoring the metadata would have called a solvent lender bankrupt. Form types are trusted, with one test attached: a Form 25 or Form 15 counts as a company exit only when no 10-K or 10-Q follows it, which correctly rejects the single-class delistings at Carvana, Medallion, and OppFi. DOES THE SCORE SEPARATE THE FAILURES? Partly, and the honest answer matters more than a flattering one. The 3 filers that departed carry a median peak score of 6 in their final eight reported quarters, against 2 for the 23 still filing. But CRMT scores above every departed filer while still operating, and Nicholas Financial departed with a peak of 2 because it sold its loan book rather than deteriorating through it. Three departures cannot support a lift table, a survival curve, or a probability. What follows is case anatomy: what the tape showed on the way out, published as anatomy and nothing more. CONN peak 8 · departed CURO peak 6 · departed NICK peak 2 · departed CRMT peak 10 · live WRLD peak 7 · live ATLC peak 6 · live HAPN peak 6 · live PRG peak 4 · live SYF peak 4 · live THE FLAG SET FLAG FIRES WHEN WEIGHT FILERS NOW equity_negative Shareholders' equity is negative 4 1 scissors Allowance coverage falling while the loss rate rises 3 0 reserve_bleed_deep Provision below net charge-offs for four or more quarters 3 1 dq_vs_reserve Past due rising while allowance coverage falls 3 0 nonreliance Filed a non-reliance 8-K 3 1 reserve_bleed Provision below net charge-offs for consecutive quarters 2 3 growth_thin Book growing while reserve coverage thins 2 0 dq_rising Past-due share climbing year over year 2 0 reserve_thin Allowance covers under three quarters of the current loss burn 2 3 equity_eroding Equity share of assets falling sharply year over year 2 6 coverage_falling Allowance coverage down year over year 1 3 losses_rising Net charge-off rate up year over year 1 1 nonaccrual_rising Nonaccrual share climbing 1 0 equity_thin Equity under a tenth of assets 1 3 late_filing Filed a late-filing notice 1 0 book_disposed Receivables book no longer reported 0 0 The composite score is the sum of active flag weights, capped at 10. It ranks deterioration signatures. It is not a probability of failure and nothing here should be read as one. WHAT IS DELIBERATELY MISSING Delinquency bucket tables live in dimensioned XBRL that the companyfacts feed does not carry. Where a filer tags an undimensioned past-due aggregate it is shown and charted; where none exists the column is blank rather than approximated. Lease-to-own filers and Carvana hold no comparable loan book, so their credit ratios stay empty on purpose rather than being invented from trade or lease receivables. The loan-level view of what servicers actually do to a delinquent account lives next door in the ABS surveillance room. THE DEAL ROOM'S SYSTEM What this is. The true sibling of the credit union tape room. NCUA's census unit is the institution; here it is the securitization, because that is the unit the disclosure regime standardizes. Each monthly EX-102 asset file carries every loan's balance, days past due, extension count, charge-off and recovery, which is richer than anything NCUA publishes, for the deals the regime covers. The band is seasoning, not calendar. A deal six months old and a deal thirty months old share nothing useful in the same calendar month, so every median here is computed across deals at the same age in months since first tape. That is the honest analogue of the credit union band median. Where only one deal has reached a given age, no band is drawn and the read says so instead of implying a comparison. The flagship analytic. Servicers grant payment extensions that return delinquent loans to current status. Reported delinquency falls; nothing about the borrower has changed. This room rebuilds the delinquency series with recently extended loans added back over a six-month lookback, and tracks how many extended loans return to 60+ within six months. The method and its limits are documented in the pipeline's METHOD.md. What a pool is not. A securitized pool is a selected slice of the sponsor's originations, subject to eligibility criteria and survivorship. Deal metrics describe the deal, and say something about the sponsor only in trend and only carefully. Reg AB II carries no dealer identifier either, so source-level attribution is impossible here by design. No presale bands loaded. A cumulative net loss number means most against the rating agencies' presale expected-loss range. Those are hand-entered per deal from the presale report and none is loaded yet, so CNL is compared only against the peer seasoning band and every read says which comparison it is using. THE FLAG SET FLAG FIRES WHEN WEIGHT DEALS NOW masking_severe Extension-adjusted 60+ delinquency exceeds reported by more than 200bps 3 17 masking Extension-adjusted 60+ delinquency exceeds reported by more than 75bps 2 0 ext_stock_rising Share of the pool touched by an extension rising sharply 2 0 cnl_above_band Cumulative net loss above the peer median at the same seasoning 2 19 dq_above_band Reported 60+ delinquency above the peer median at the same seasoning 1 18 recovery_weak Cumulative recoveries under a third of charged-off principal 1 12 DEALS REJECTED BEFORE PUBLICATION A tape assembled while the disk was full silently loses months, and a deal missing months understates cumulative loss and pool factor in ways that look like good performance. Any calendar gap between the first and last tape disqualifies a deal from this room. DEAL MONTHS PARSED GAPS AMCAR-2022-1 49 2024-02-29 -> 2024-04-30; 2024-06-30 -> 2024-08-31; 2024-11-30 -> 2025-02-28 EART-2021-1 55 2021-11-30 -> 2022-01-31; 2022-12-31 -> 2023-02-28 GMCAR-2022-1 49 2022-10-31 -> 2022-12-31; 2023-01-31 -> 2023-03-31 WOSAT-2023-A 40 2024-02-29 -> 2024-04-30; 2025-12-31 -> 2026-03-31 REDEFAULT COHORTS: EXTENDED LOANS BACK TO 60+ WITHIN SIX MONTHS EXTENSION MONTH DEAL LOANS EXTENDED REDEFAULT 6M MONTHS OBSERVED 2026-03-31 WOSAT-2023-A 1 0.0% 4 2025-01-31 WOSAT-2023-A 1 0.0% 6 2024-11-30 WOSAT-2023-A 1 0.0% 6 2024-06-30 WOSAT-2023-A 1 0.0% 6 2024-04-30 WOSAT-2023-A 4 0.0% 6 2023-10-31 WOSAT-2023-A 2 0.0% 6 2023-07-31 WOSAT-2023-A 4 0.0% 6 2023-03-31 WOSAT-2023-A 2 0.0% 6 2023-02-28 WOSAT-2023-A 1 0.0% 6 2026-03-31 WOART-2025-A 1 0.0% 4 2026-01-31 WOART-2025-A 1 0.0% 6 2025-10-31 WOART-2025-A 1 0.0% 6