# Caliber Collision — Post-Repair ADAS Evidence Pilot

**Pilot ID:** `caliber-collision-evidence-2026`  
**Mode:** Shadow scoring → signed evidence at structural release  
**Duration:** 10–12 weeks · **Regions:** 3–5 MSO regions · 4–8 centers each

---

## Executive summary

Caliber Collision is a national MSO repairing thousands of ADAS-equipped vehicles monthly. Structural repairs — bumper, fender, rail, sensor mount — shift camera and radar extrinsics. OEM programs and insurers increasingly require **documented calibration validation**, not just a scan printout in a folder.

NADIR provides:

1. **Continuous residual scoring** after repair close (shadow, then integrated)
2. **Signed evidence bundles** tied to repair order and tier timeline
3. **Regional consistency dashboards** — compare CRITICAL rates and calibration discipline across centers

This pilot aligns with Calibration Lab **`collision`** physics (impulse extrinsics at t≈2s) and the NADIR Console evidence export workflow.

---

## Problem statement

| Pain | MSO impact |
|------|------------|
| Uneven calibration quality across centers | OEM audit findings, brand risk |
| Insurance supplement disputes on ADAS line items | Revenue leakage, cycle time |
| No portable signed evidence at release | Re-work, customer complaints |
| Hidden drift after “successful” static scan | Liability exposure post-pickup |

Caliber wins by proving **national consistency** with audit-grade exports — not just buying more scan tools.

---

## Proposed scope

### Phase 1 — Shadow + regional dashboard (weeks 1–6)

- 3 pilot regions, 4–8 centers each (24–48 centers total)
- Ingest post-repair residuals or scan proxy on ADAS-flagged structural jobs
- Shadow tier scoring — no change to release checklist
- Regional ops dashboard: CRITICAL rate per 100 repairs, center ranking, drift timeline samples
- Compare vs matched pre-pilot baseline quarter

### Phase 2 — Evidence at structural release (weeks 7–12)

- Integrate signed NADIR evidence into release workflow (API + console)
- Target **≥92%** evidence attachment before customer pickup
- Insurer DRP sample review (n≥25) for acceptance rate metric
- OEM validation sample at region level

### Out of scope (pilot)

- National mandate across all Caliber centers
- Replacing OEM scan hardware
- Autonomous hold on vehicle release (manual review only)

---

## Evidence workflow

```
Structural repair complete → calibration performed → NADIR ingest
        → tier score + drift timeline → signed evidence bundle
        → attached to RO → insurer/OEM audit trail
```

Evidence includes:

- Drift residual timeline (pre/post repair)
- Tier transition log (NOMINAL → CAUTION → CRITICAL)
- Sensor fault vector (camera °, radar mm, LiDAR mm)
- Recommended action + Merkle-signed export hash

Demo: `/console#evidence` + `/evidence` product page.

---

## Calibration Lab alignment

**Collision impulse** scenario in Calibration Lab:

- Mount shock at **t = 2.0s** post-impact
- Camera + radar + LiDAR coupled drift
- Auto-benchmark ranks detection algorithms vs ground truth

Use in regional technician training and insurer conversations.

---

## Success metrics

Full definitions: [`success_metrics.json`](success_metrics.json)

| Metric | Target |
|--------|--------|
| Signed evidence at structural release | ≥ 92% |
| Cross-center tier variance reduction | ≥ 20% |
| Post-collision drift detection (48h) | ≥ 88% |
| Insurer evidence acceptance | ≥ 80% |
| OEM audit pass proxy | ≥ 85% |

---

## Data requirements

See [`data_requirements.md`](data_requirements.md).

---

## Deliverables

| Week | Deliverable |
|------|-------------|
| 0 | Region/center selection + baseline KPI pull |
| 2 | Shadow ingest + regional dashboard live |
| 6 | Phase 1 metrics vs variance target |
| 8 | Insurer sample pack (5 bundles) |
| 12 | MSO rollout proposal + pricing |

---

## Commercial path

| Stage | Model |
|-------|--------|
| Regional pilot (3 regions) | $2.8k/mo + per-center ingest |
| MSO platform | Volume tiering by repair count |
| Insurer API | Evidence delivery to DRP partners |

---

## Related pilots

- **Safelite** (glass QA): `../safelite/` — complementary, not competitive
- **Amazon** (fleet): `../amazon-logistics/` — fleet angle for insurer narrative

---

## LOI

Custom exhibit: [`loi_exhibit_a.md`](loi_exhibit_a.md)