Deep dive — chart 3

Chart 3 — lane rate per borrower group, baseline vs recent

The chart with the operational + regulatory bite. Where risk-score distribution shifts translate into actual credit decisions. B3 decline / secondary-market rate rose 18% → 62%. Other groups stable. That is the exact adverse-impact signature fair-lending reviewers look for.

Underwriting lane rate per borrower group, baseline vs recent

Underwriting lane rate per borrower group — baseline (left) vs recent (right). B3 decline / secondary-market rate rose sharply while other borrower groups stayed stable.

What you are looking at

Grouped bars per borrower group (B1 / B2 / B3 / B4 / B5)
Three colors per borrower group: Auto-Approve Prime (green) / Manual Review / Price-Up (blue) / Decline / Secondary Market (red)
Left cluster: baseline (first 30 days) — the "normal" routing per borrower group
Right cluster: recent (last 30 days) — the current routing per borrower group

What the chart shows

B1, B4, B5 — lane rates roughly unchanged between baseline and recent.
B2 — modest decline-lane rise consistent with the KL-divergence signal.
B3 — decline / secondary-market rate rises sharply (18% baseline → 62% recent). Auto-Approve share shrinks. Manual Review absorbs some of the shift.
The dramatic differential exists only for B3.

44 percentage-point differential in decline / secondary-market routing. One borrower group. Silently. Over 45+ days. This is the exact operational pattern the CFPB adverse-action-notice supervisory question and the ECOA / Reg B disparate-impact analysis are calibrated to detect.

Why aggregate metrics missed it

Because B3 is one borrower group among five, the aggregate approval rate barely moved: the B3 surge into decline / secondary-market rounded to noise across the whole portfolio. That is why the model-risk-management dashboard stayed green. Aggregate metrics are inherently blind to group-differential patterns, and group-differential patterns are inherently what disparate-impact analysis is built to surface.

The signal that fired the A/B/C determination

Across the 90-day audit period, the sensor's per-group KL-divergence signal fired for B3 on Days 44, 51, 58, 65, 72, 79, and 86. Repeated, sustained, and high-severity. Once repeated group-differential signals fire, the Snapshot lands on Category C by rule — no matter what any aggregate metric shows.

See the A / B / C decision logic →

The exhibit-quality read

This chart is what a plaintiff's expert witness or state AG fair-lending investigator asks the bank for first. It is also what the bank's General Counsel should have on file before either lands — because the same chart, in the bank's hand, tells a different story:

Same chart, bank stance: "we identified this pattern through independent verification on Day X, initiated the following remediation on Day X+N, and here is the timeline of that remediation." The chart becomes evidence of good-faith investigation, not evidence of missed monitoring.

$499 Snapshot. 3 business days.

Same lane-shift chart for your bank's actual AI credit-underwriting surface + full determination + 3 fix-first items.

Buy $499

Back to banking sample hub.