Automotive WIP tracking — what production-grade actually looks like.

Why automotive WIP is the hardest UWB design problem.

Body shops generate the worst multipath conditions in industrial RTLS.

Large flat metal surfaces (vehicle bodies, press tooling, stillage), stainless production guarding, concrete-aggregate floors, overhead conveyor and dynamic metal movement through the cycle — every one of them creates RF reflections.

The vendor reference design absorbs this by adding anchors; independent design absorbs it by geometric offset.

Real production-grade UWB in automotive body-in-white hits 30 cm @ 95th percentile across the shift, not just at commissioning. The design that gets there is anchor-count-disciplined, height-verified, multipath-aware and acceptance-tested at the press cycle peak.

The KPI stack that decides ROI.

WIP accuracy: 99.2% across three shift patterns is what production-grade looks like. Below 95% the OEE uplift evaporates.

Sequence integrity: for JIS lines, every torque event, every fitment, every off-line repair tagged to VIN. IATF 16949 audit-grade.

Mis-build rate: typical baseline 0.4–0.8% of vehicles; production-grade RTLS-tied traceability drives this to under 0.1%.

OEE points captured: 6–9 points typical at six months post go-live across our automotive client base.

The technology stack we recommend most often.

UWB on the line for sub-30 cm WIP position (Sewio, Ubisense, Pozyx, depending on procurement constraints).

Passive RFID at portals for cross-line, stillage and dock-door reads (Impinj, Zebra fixed readers).

BLE-AoA in support zones (paint, sub-assembly) where UWB density would be excessive.

Integration: SAP MII, Rockwell FactoryTalk, Siemens Opcenter, or your in-house MES — the event stream feeds existing andon, downtime tracking and quality genealogy.

Where the cost actually lives.

Hardware on a 5,000 m² body shop: UWB anchors €200–€300 each; 40–60 of them; €12,000–€18,000 hardware.

Install on the same site: €1,000–€1,500 per anchor including RF survey, mount, sync calibration, commissioning under production load. €40,000–€90,000 install.

Integration into MES: typically 20–40% of total programme cost. The line item that decides whether adoption actually sticks.

Independent vendor-design audit (Engagement C in our install service) typically saves €20,000–€50,000 install per site, scaling across multi-site rollouts of 5–15 plants.

UWB RTLS vendors (Sewio, Ubisense, Litum, Pozyx and peers), Passive RFID SI packages, and MES analytics overlays all compete for WIP visibility budgets. Body and paint shops are hostile RF environments — metal, movers, pits — so glossy warehouse demos transfer poorly. Hybrid designs win: Passive RFID at docks and line gates for cheap presence events; UWB where continuous in-flight position drives andon, sequencing or buffer control.

What programmes get wrong: UWB-blanket coverage when portals would have paid back faster; ignoring vehicle/carrier identity that must survive paint and weld; measuring localisation centimetres instead of dwell, starve and rework loops; and letting an AMR vendor redefine WIP tracking as robot fleet monitoring.

We remain vendor-neutral and sell no radios or robots. Design centres on the KPI stack (dwell, buffer health, sequence integrity), RF survey under live production, and MES event contracts. See JIS when sequence is the dominant job.

Automotive WIP is a hostile RF and process environment: moving metal, pits, carriers through paint, and shift patterns that change sightlines hourly. Designs copied from tidy warehouses fail. Start from the KPI stack — dwell by station, buffer health, rework loops, starve risk — then place Passive RFID at gates and UWB only where continuous position changes a decision.

Carrier and body identity must persist through processes that destroy ordinary tags. Mounting, shielding and process-tolerant encoding are engineering problems, not purchasing afterthoughts. MES integration should drive andon and sequencing logic from validated events; a parallel RTLS dashboard operators ignore is waste. Live-shift RF surveys across paint, body and assembly catch multipath the empty-bay survey never sees.

AMR and AGV programmes should consume the same WIP truth layer. Buying robots to compensate for missing inventory truth usually produces expensive choreography on bad data — we advise and integrate; we sell no fleets. Change control when lines rebalance is mandatory: yesterday’s anchor map is today’s accuracy regression.

Acceptance: KPI movement on dwell and starve measures, sustained identity continuity through the worst process step in scope, and a re-validation ritual after every major layout change.

UWB RTLS vendors (Sewio, Ubisense, Litum, Pozyx and peers), Passive RFID SI packages, and MES analytics overlays all compete for WIP visibility budgets. Body and paint shops are hostile RF environments — metal, movers, pits — so glossy warehouse demos transfer poorly. Hybrid designs win: Passive RFID at docks and line gates for cheap presence events; UWB where continuous in-flight position drives andon, sequencing or buffer control.

What programmes get wrong: UWB-blanket coverage when portals would have paid back faster; ignoring vehicle/carrier identity that must survive paint and weld; measuring localisation centimetres instead of dwell, starve and rework loops; and letting an AMR vendor redefine WIP tracking as robot fleet monitoring.

We remain vendor-neutral and sell no radios or robots. Design centres on the KPI stack (dwell, buffer health, sequence integrity), RF survey under live production, and MES event contracts. See JIS when sequence is the dominant job.

Automotive WIP is a hostile RF and process environment: moving metal, pits, carriers through paint, and shift patterns that change sightlines hourly. Designs copied from tidy warehouses fail. Start from the KPI stack — dwell by station, buffer health, rework loops, starve risk — then place Passive RFID at gates and UWB only where continuous position changes a decision.

Carrier and body identity must persist through processes that destroy ordinary tags. Mounting, shielding and process-tolerant encoding are engineering problems, not purchasing afterthoughts. MES integration should drive andon and sequencing logic from validated events; a parallel RTLS dashboard operators ignore is waste. Live-shift RF surveys across paint, body and assembly catch multipath the empty-bay survey never sees.

AMR and AGV programmes should consume the same WIP truth layer. Buying robots to compensate for missing inventory truth usually produces expensive choreography on bad data — we advise and integrate; we sell no fleets. Change control when lines rebalance is mandatory: yesterday’s anchor map is today’s accuracy regression.

Acceptance: KPI movement on dwell and starve measures, sustained identity continuity through the worst process step in scope, and a re-validation ritual after every major layout change.