Sustainability bought the sensors. Quality never asked for the feed.

The hydraulic press started drawing four per cent more current on a Tuesday. I know the number precisely because I went back afterwards and looked, which is the embarrassing part – the trend had been sitting in the energy dashboard all week, plain as a tram line. Scrap spiked on the Friday. Nobody compared the two screens until the 8D was half-written, containment had eaten a pallet of parts, and a customer had rung me personally. The earliest signal in that plant was electrical. We ignored it for one reason: it lived on somebody else's dashboard.

The trade press this month reads like an instrumentation catalogue. Smart-factory write-ups celebrate energy tracking as the EV-readiness win, Battery Show Detroit is filling halls with monitoring kit, and the Bosch–Dassault Systèmes partnership promises AI-led telemetry on anything with a cable attached. Nearly all of it bought with sustainability money and owned by facilities. The observation missing from the coverage: a power meter does not know which budget paid for it. It is the cheapest real-time process signal a quality department will ever get, and I have yet to audit a control plan that mentions it.

what the meter already sees

A machine converts electrical energy into work. Anything that changes the work – tool wear, a bearing going rough, a guide running dry, incoming material drifting hard – shows up in the energy signature before it shows up in the part. Not a metaphor; arithmetic. Wear adds friction, friction consumes torque, torque draws current. Geometry tolerates wear until it suddenly doesn't. The ammeter has no such patience.

• Motor-current draw. A press, spindle or mixer pulling more amps for the same cycle is telling you the process changed. Tool wear and mechanical degradation bill you in kilowatts first, microns second.
• Heater duty cycle. A degrading band heater still holds setpoint – by staying on longer. Cure, viscosity and fill behaviour drift before any dimension does.
• Compressed-air signatures. A cylinder with a tired seal strokes differently and leaks more. Consumption patterns shift long before the force–displacement curve catches it.

A decent power meter on a machine circuit costs a few hundred euros. A coordinate measuring machine starts at six figures. One tells you the process is drifting this Tuesday; the other tells you it drifted last week.

A machine announces its intentions in amperes. Most quality departments read the obituary in the scrap report.

two departments, one plant

I watched this silo form on day one of a greenfield build – a 900-plus-employee plant where I created the QA/QC department from the ground up. You would expect a greenfield, with no legacy systems and no legacy grudges, to get data architecture right. It doesn't. Facilities commissioned their energy dashboard, quality commissioned its own views, the same controllers fed both, and not one tag was shared. By week three a greenfield has legacy. Nobody owns the joins between dashboards, so nobody defends them, so they close.

The bill arrived in the QRQC room. Our 70 per cent reduction in defect costs at that plant came from reacting to the earliest signal instead of the finished nonconformance – not from cleverer statisticians. But QRQC can only react to what it can see. If the earliest signal sits behind a facilities login, your rapid-response discipline is structurally handicapped. At WITTE Automotive the Q-Wall existed for exactly this: put weak signals in front of people hourly, before they graduate into strong ones. The wall is only as good as what you are allowed to pin to it.

Reading an ammeter before the defect is the same craft as reading security logs before the breach. Different dashboard, identical discipline: believe the weak signal or explain the strong one later, with a customer on the line.

turning telemetry into a control

None of this needs new infrastructure, which is the point. The plant has already paid for the meters, the historians and the OPC-UA pipes; quality needs read access and a bit of discipline. Start by asking facilities for the feed, not the dashboard – they keep their KPIs, you get the tags in your SPC views, and nobody fights a turf war. Then set trial limits from capability runs: log kW, duty cycle or air consumption across a known-good batch, compute the natural spread, set provisional control limits, and validate. If the tag moved before your last three 8Ds opened, promote it. If it correlates with nothing, retire it – honest triage keeps the channel credible.

When a tag moves, escalate it through QRQC like any failed dimension: stop, contain, diagnose at the source. And write the channel into the control plan and PFMEA as a leading indicator with a defined reaction. That is what survives audits, reorganisations and the departure of the one engineer who believed in it. IATF 16949 auditors respond well to controls with evidence trails; they respond badly to signals living in someone's holiday-cover spreadsheet.

Everyone instrumenting a plant for energy reasons believes they bought a sustainability tool. They bought a quality instrument and haven't noticed. The cheapest capability upgrade available this quarter is not a sensor, a system or a consultancy – it is a phone call to facilities asking for read access to data your own machines already produce. Across two decades in automotive and aerospace I have not seen a cheaper signal with a better lead time. When the meter twitches, believe it. The defect is already on its way; the only question is whether you read about it on Tuesday or explain it on Friday.