# Safelite — Post-Glass ADAS Calibration QA Pilot

**Pilot ID:** `safelite-glass-qa-2026`  
**Mode:** Shadow QA + evidence-at-close (optional Phase 2 gate before release)  
**Duration:** 8–12 weeks · **Markets:** 5–25 locations

---

## Executive summary

Safelite replaces more windshields in North America than any other provider. A growing share of those jobs include **forward-facing ADAS cameras** requiring static or dynamic calibration after glass R&R. Inconsistent calibration quality drives comebacks, fleet customer complaints, OEM audit findings, and insurance disputes.

NADIR provides **continuous residual scoring** and **signed evidence bundles** that answer one question at repair close: *Is this vehicle's ADAS stack aligned enough to release — with proof?*

This pilot pairs Safelite's service workflow with NADIR's glass-misalignment physics model (Calibration Lab `glass` scenario) and post-repair evidence exports.

---

## Problem statement

| Pain | Business impact |
|------|-----------------|
| Hidden camera pitch/yaw error after glass install | AEB / LKAS under-performance |
| Uneven calibration discipline across locations | Brand risk with fleet OEM partners |
| No portable evidence at work-order close | Claims disputes, audit gaps |
| Comeback recalibrations | Bay capacity loss, customer friction |

**Industry context:** Post-repair ADAS calibration is moving from optional to mandatory across OEM and fleet programs. Safelite is positioned to **set the national QA standard** with measurable evidence.

---

## Proposed scope

### Phase 1 — Shadow QA (weeks 1–6)

- 5–12 pilot locations in distinct DMAs
- Ingest post-repair telemetry or scan proxy after windshield jobs flagged ADAS-equipped
- NADIR scores tier **in shadow** — no change to release process
- Compare NADIR tier vs existing QA checklist and comeback tickets
- Weekly QA digest: locations with elevated CRITICAL rate, sample evidence bundles

### Phase 2 — Evidence at close (weeks 7–12, optional)

- Integrate NADIR evidence export into work-order close (API or batch)
- **Gate:** only locations with ≥80% technician adoption in Phase 1
- Target: 95% of ADAS glass jobs ship with signed NADIR evidence attached

### Out of scope (pilot)

- Replacing OEM scan tools
- Mandating recalibration on all NOMINAL tiers
- National rollout commitment (conversion subject to success criteria)

---

## Calibration Lab alignment

Safelite pilot maps directly to NADIR Calibration Lab damage scenario **`glass`** (service misalignment):

- Onset at **t ≈ 0.5s** after windshield R&R
- Camera pitch/yaw step + radar coupling
- Demo link: `/calibration-lab` → select **Glass / service misalignment**

Use in sales conversations and technician training: *"This is the physics model we score against your post-repair ingest."*

---

## Success metrics

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

| Metric | Target |
|--------|--------|
| Post-glass drift detection (24h) | ≥ 90% |
| Calibration comeback reduction vs control | ≥ 25% |
| Evidence bundle at repair close | ≥ 95% |
| Unnecessary recalibration rate | ≤ 8% |
| Tier agreement with OEM scan (sample) | ≥ 85% |

---

## Data requirements

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

Core feeds:

- Work order ID, location, timestamp, ADAS flag, calibration performed (Y/N)
- Pre/post scan results where available
- Optional: fleet customer ID for B2B reporting
- Comeback / rework tickets within 30 days (ground truth)

---

## Deliverables

| Week | Deliverable |
|------|-------------|
| 0 | Location selection + technician training (30 min + Lab demo) |
| 2 | Shadow ingest live; first QA heatmap by location |
| 6 | Phase 1 metrics vs success criteria |
| 8 | Phase 2 integration plan (if proceeding) |
| 12 | Final report + national rollout pricing proposal |

---

## Commercial path

| Stage | Pricing guidance |
|-------|------------------|
| QA pilot (5–12 locations) | $2.8k/mo platform + training |
| Regional expansion | Per-location/month tiering |
| National MSO | Volume agreement + API SLA |

---

## LOI customization

Safelite-specific clauses in [`loi_exhibit_a.md`](loi_exhibit_a.md):

- Windshield R&R as primary ground-truth event type
- QA ownership on Safelite side; NADIR provides scoring + evidence only
- Fleet customer reporting addendum (optional exhibit B — commit 22)

---

## Related assets

- Amazon design-partner package (fleet angle): `../amazon-logistics/`
- Article: [/articles/post-repair-adas-calibration](/articles/post-repair-adas-calibration)
- LOI master template: `NADIR_DOCUMENTATION/09_execution_plans/loi_template.md`