Repair RO exports (CSV or JSON) plus telematics JSONL or a fleet API stream. We define a minimal contract at kickoff: VIN, timestamps, and the repair close anchor for MTTFF-RE.
Drift happens quietly.
We make it obvious.
Post-repair ADAS health monitoring for commercial fleets — running in shadow mode on the telematics you already export.
01 / 06 Connect existing telematics Read-only Geotab link discovers 118 ADAS assets and backfills 30 days of history. Walkthrough page ›
Works with the stack you already run
Geotab
Samsara
Motive
Verizon Connect
Ford Pro
GM Fleet
Mobileye
CSV · JSONL · REST
Integration names indicate compatible data paths, not endorsements or partnerships.
The post-repair gap
A clean scan is not a straight sensor.
Roughly two-thirds of collision repair orders touch ADAS components, and a significant share of required recalibrations are skipped or left incomplete. A static bay check can pass while the camera still points a fraction of a degree off — which becomes metres of error at automatic-braking range.
~65%
of collision repair orders touch ADAS-related components.
Industry estimate
~35–40%
of required recalibrations are skipped or incomplete.
Industry estimate
≤7 days
median time to first flag after a repair event — the pilot target.
MTTFF-RE target
0
ECU writes. Detection and evidence only, never actuation.
By design
The platform
Four modules. One contract.
From the moment a repair order closes to the evidence bundle an insurer accepts.
Monitor
Score every sensor frame against the road, not the bay.
Repair-order webhooks and telematics batches feed cross-modal residual scoring, tier transitions, and full history per VIN.
Explore MonitorOperate
A queue your maintenance team can actually work.
CRITICAL vehicles surface in Console with a weekly digest, owner assignment, and no spreadsheet running alongside it.
Explore OperateProve
Evidence that survives an audit.
Signed evidence bundles with HMAC-SHA256 lineage, model version, and MTTFF-RE reporting — exportable for insurers and safety boards.
Explore ProveResolve · Roadmap
Route the flagged vehicle to a bay that can fix it.
Partner calibration routing, scheduled honestly as roadmap work rather than shipped capability.
Explore ResolveHow it works
Four steps a fleet IT reviewer can follow in one sitting.
01
Connect
Repair orders arrive by CSV, JSONL, or webhook, alongside the telematics you already export. No new hardware.
02
Score
Cross-modal residuals run in shadow mode and assign a health tier to every vehicle, anchored to the repair close.
03
Queue
CRITICAL vehicles surface in Console with an owner and a weekly digest — before they return to route.
04
Prove
A signed evidence bundle documents the detection, the lineage, the model version, and the outcome.
Integration
Two endpoints and the export you already run.
Monitor needs repair-order close timestamps joined to VIN, plus the telematics feed your fleet already produces. Post them over REST, call the Python SDK from an existing job, or run the Pulse edge pack when the data is not allowed to leave your network. There is no new hardware in the vehicle and no agent on the ECU.
- Telemetry
POST /v1/telemetry- Repair events
POST /v1/repair-events- SDK
- Python
nadir_sdk - On-premise
- Pulse edge pack
// A repair order closing is the anchor for MTTFF-RE.
{
"vin": "1FUJGLDR9CSSE4417",
"ro_number": "RO-48812",
"closed_at": "2026-07-21T16:04:00Z",
"operations": ["WINDSHIELD_REPLACEMENT"],
"odometer_km": 184002
}
// Scoring runs in shadow mode. Nothing is written back.
{
"event_id": "evt_8fd21c",
"tier": "CAUTION",
"residual_sigma": 0.42,
"first_flag_hours": 101.5
}
Why residuals
Millimetres at the bracket. Metres at braking range.
A camera rotated half a degree still produces a clean image and a passing self-test. What changes is where the vehicle believes objects are. NADIR compares what each sensor claims about the road against what the others claim, then reports the disagreement — never the guess.
Safety boundary
A monitoring system that cannot touch the vehicle.
NADIR is deliberately confined to detection and evidence. It reads data, scores disagreement between sensors, and reports it. It does not steer, brake, calibrate, or write to an ECU — which is what allows it to be deployed on a working fleet without entering the functional-safety path.
- Actuation
- None
- ECU writes
- Zero, by design
- Classification
- Shadow advisory · not ASIL
- Correction
- Partner bay · human calibration
Built for the whole fleet
One metric. Every vehicle. Every week.
MTTFF-RE — mean time to first flag after a repair event — is the single number a pilot is judged on. Median under seven days, with at least 70% of flagged events inside target. Everything else in the platform exists to make that number honest.
Four-week shadow proof
$750 to find out whether your fleet has a drift problem.
Fifty to a hundred and fifty vehicles, advisory only, no ECU writes, no rip-and-replace. Week four closes with the MTTFF-RE readout and an exported evidence bundle you keep either way.
- Weekly CRITICAL digest from week one
- Console access for your maintenance leads
- MTTFF-RE readout against the 168-hour target
- Signed evidence bundles you keep either way
- Named engineer on a shared channel
- Pilot size
- 50–150 vehicles
- Duration
- 4 weeks, shadow mode
- Data in
- Repair ROs + telematics
- Data out
- Signed evidence bundles
Evidence
Bundles that survive the standards your auditors cite.
Every flag produces an evidence bundle: canonical JSON carrying the residual lineage, the model version that produced the score, the repair event it is anchored to, and an HMAC-SHA256 signature over the payload. It is designed to be handed to an insurer, a safety board, or a customer's own QA without interpretation.
- Format
- Canonical JSON
- Signature
- HMAC-SHA256
- Lineage
- Residuals + model version
- KPI target
- ≤168h on ≥70% of events
Standards mapping
Written against the regulations fleet safety teams are already being measured on.
FMVSS 127
UN ECE R152
ISO 26262
Questions
Clear answers before you connect anything.
Mean time to first flag after repair event — hours from RO close to first CAUTION or CRITICAL tier. Pilot success: median ≤ 168 hours and ≥ 70% of flagged events within target.
No. NADIR detects likely miscalibration and documents evidence for inspection. Mechanical damage, installation faults, and unsupported conditions still require qualified physical calibration in the bay.
Never. NADIR is advisory analytics. It reads exported data and returns scores, queues, and evidence. There is no path in the product that writes to an ECU or modifies a calibration store.
Once RO import and telematics paths are agreed, shadow scoring typically starts within days. Week four closes with the MTTFF-RE readout and evidence bundle export.
Yes. SDKs, REST APIs, model-version metadata, and signed evidence bundles let your team trace every score and repair-event join back to source data.
Know before they
return to route.
Post-repair detection. Shadow mode. One metric. Signed proof.