Deep dive — chart 2

Chart 2 — B3 baseline vs recent distribution

The how. Where Chart 1 shows when the B3 line drifted upward, this chart shows how the B3 risk-score distribution shape shifted. Mass moved right, into the decline / secondary-market lane.

B3 baseline vs recent underwriting risk-score distribution comparison

B3 risk-score distribution — baseline (blue) vs recent (orange). Mass shifted right, past the decline / secondary-market threshold at score 67.

What you are looking at

X-axis: underwriting risk-score (0 to 100, higher = riskier)
Y-axis: density of applications at that score for B3
Blue distribution: baseline window (first 30 days), what B3 scoring normally looked like
Orange distribution: recent window (last 30 days), what B3 scoring looked like at Snapshot time
Vertical line at score 67: threshold for routing to decline / secondary-market lane

What the shape shift means

The orange distribution's mass has moved right compared to the blue distribution. Three things are happening simultaneously:

The peak (mode) climbs. The most common B3 score used to sit in the Manual Review / Price-Up band; now it sits at or above the decline threshold.
The left tail shrinks. Fewer B3 applications score in the Auto-Approve Prime band.
The right tail grows. Many more B3 applications score above the decline / secondary-market threshold.

Credit-decision translation: the AI now systematically scores the post-shift B3 applicant profile (debt-consolidation purpose, higher DTI, larger requested amounts, modestly weaker FICO-at-decision) as much riskier than baseline. The score does not distinguish between "moderate DTI, debt-consolidation, near-prime" (a legitimate credit-worthy profile) and "distressed sub-prime." The distribution shape drift is the mathematical fingerprint of that decision-model failure mode.

What KL-divergence quantifies

Kullback-Leibler divergence measures how different one distribution is from another. Values:

KL < 0.05: the same distribution (random noise)
KL 0.05-0.2: mild shift, monitor
KL 0.2-0.5: significant shape shift, flag as event
KL > 0.5: substantial shape shift, high-severity event
KL > 1.0: distributions are almost non-overlapping — the applicant population the AI is scoring is behaviorally different from the population it was trained on

The sample's B3 KL-divergence hit 0.587 on Day 44 and 0.845 on Day 72. B2 hit 1.885 on Day 86. B5 hit 1.132 on Day 86. All well above threshold, and the pattern stayed.

The credit-underwriting + regulatory read

Credit-underwriting read: the AI has drifted out of calibration for the current B3 applicant profile. Manual underwriter re-review should apply until retraining + validation.

Regulatory read: the shape shift is precisely the kind of "measurable drift in performance on a specific subpopulation" that SR 11-7, OCC Bulletin 2011-12, and CFPB Circular 2022-03 expect the bank to detect and act on.

The next chart

The baseline-vs-recent chart shows how the B3 distribution shifted. The lane-rate chart shows the operational consequence: which lane are B3 applicants getting routed to now vs baseline.

Chart 3 — lane rate per borrower group →

$499 Snapshot. 3 business days.

Same shape-analysis for your bank's actual AI credit-underwriting surface + full analysis report.

Buy $499

Back to banking sample hub.