# CU Intelligence · System

the system by asset band, the validation lift table, the deterioration signature into departure, score distribution.

Live page: https://getbookiq.com/cu-intel/#system

---

THE SYSTEM AT 2026-03
Vehicle share of the loan book and dealer-sourced share invert with size: the small end writes car loans, the large end buys them. Median staffing is the capacity ceiling: below ~$500M nobody can dedicate an analyst.
BAND	CUS	ASSETS $B	VEH %LOANS	IND %VEH	MEDIAN FTE	MEDIAN DQ	MEDIAN SCORE	SCORE 5+
<$50M	1,931	33.9	52.7%	5.7%	3	0.66%	3	21.4%
$50-100M	577	41.8	42.7%	19.2%	13	0.63%	2	10.2%
$100-200M	547	78.5	37.3%	35.6%	27	0.62%	2	9%
$200-500M	523	169.4	34.2%	52.5%	60	0.6%	2	5.9%
$500M-1B	284	204.8	30.2%	61.4%	123	0.59%	2	10.2%
$1-10B	450	1303.6	28.1%	83.3%	308	0.62%	2	9.6%
>$10B	24	675.7	22.4%	62.6%	2243	0.91%	2	4.2%
VALIDATION: DEPARTURE RATE WITHIN 4 QUARTERS, BY RISK SCORE
Base rate 4.33%, 1749 departures on our watch. The claim is lift, not that every flagged institution is dying: most departures are voluntary mergers.
0-1
1.82% · 0.4x base
2
2.74% · 0.6x base
3-4
4.33% · 1.0x base
5-6
9.1% · 2.1x base
7+
22.97% · 5.3x base
THE DETERIORATION SIGNATURE · MEDIAN PATH INTO DEPARTURE
QUARTERS TO DEPARTURE	NET WORTH %	ROA %	DELINQUENCY %	MEMBERS YOY %	SCORE
t-7	11.24	0.14	0.65	-1.99	3
t-6	11.25	0.11	0.68	-2.04	3
t-5	11.17	0.1	0.74	-2.19	3
t-4	11.21	0.09	0.73	-2.17	3
t-3	11.19	0.06	0.7	-2.36	4
t-2	11.1	0.04	0.74	-2.79	4
t-1	10.95	0	0.72	-3.27	4
t-0	10.67	-0.44	0.66	-4.63	5
survivors (pooled)	11.44	0.55	0.55	0	2
CONVENTIONS & HONESTY
Source: NCUA 5300 bulk files, resolved per cycle through the account dictionary. Ratios use period-end balances and YTD annualization (Q1×4, Q2×2, Q3×4/3), so ROA and yields land within a few bp of NCUA's average-balance FPR math: screening-grade, not examiner-identical. The bulk data lags NCUA's single-CU FPR site by roughly one quarter. Indirect figures carry known self-reporting artifacts; filers tripping reclassification checks are labeled on their tape read. A flag is a question to ask management, not an accusation.