A 99.9% battery yield isn't quality. It's a safety incident on a delay.

I stood in a battery cell plant recently — a good one, certified, proud of its scrap rate — and watched a quality engineer present a yield improvement dashboard. 99.93%. Applause. I asked one question: of the 0.07%, how many cells passed final QC and shipped? He didn't have the number. Nobody in the room did. That gap between the dashboard and the answer is where people eventually get hurt.

The trade press celebrates battery manufacturers adapting to higher standards. Read past the headlines and you find the real driver: standards are tightening because field incidents are accelerating. These two facts live in tension. Right now the yield mindset is winning when it shouldn't be in the room.

The yield trap — when your quality metric normalises the defect that propagates thermally

I've run volume production under IATF 16949. At SNOP, I built a greenfield QA/QC organisation for over 900 people from bare concrete up, and we drove 70% defect-cost reduction in the early cycles. That was the correct framework for the correct product. Stamping defects, weld spatter, assembly misalignment — visible, containable, non-catastrophic. A bad steering rack gets caught at the customer's dock. It costs money. It costs reputation. Nobody dies.

A lithium-ion cell with a separator thickness deviation doesn't fail at the dock. It passes formation, ages normally, gets built into a module, ships in a vehicle or a grid storage array, and then — fourteen, eighteen, twenty-four months later — a charge cycle triggers internal shorting. Thermal runaway. The failure mode your yield dashboard normalised at 0.07% becomes the centre of an investigation.

Your yield rate isn't a quality metric. It's a forecast of future incidents, expressed as a percentage.

Yield is a manufacturing efficiency metric. Safety assurance is a system-level discipline. The battery industry conflates them because the IATF toolkit makes it easy and the volume ramp rewards anything that keeps the line running.

What aerospace learned about latent defects that battery plants haven't

In aerospace, we operate under a different founding assumption: the defect that passes inspection but fails in service is the defect that defines your quality system. Not the ones you catch.

At Airbus, I led a cycle that cut EASA audit findings by 50% in a single pass. Not by tightening inspection stations. We did it by restructuring how the organisation thought about latent-event prevention — control plans, traceability architecture, failure mode documentation, the relationship between process parameters and service-life failure modes. AS9100 demands you treat quality as a safety boundary, not a cost function.

Battery cell manufacturing has aerospace failure physics — cascading, latent, thermally propagating — but runs on automotive quality tools built for defects that are visible, bounded, and line-stoppable. I work across both standards. The PFMEA for a battery cell should structurally resemble what we write for a flight-critical structural component, not what we write for an interior trim panel. In most cell plants I've walked through, it doesn't come close.

I have not yet seen a credible field-side thermal event PFMEA in any cell plant I've visited. The yield number says everything is fine. The yield number is a manufacturing metric being asked to do safety engineering work it was never designed for.

The control plan for a product that fails 18 months after shipment

This is where it gets concrete. Your control plan should be designed around the failure mode that happens after the product leaves your dock — not the one your end-of-line tester can catch.

In automotive volume production, APQP works because most failure modes surface within the production cycle or at receiving inspection. QRQC, 8D, containment — the defect is observable, the loop closes, the cost is bounded. I've deployed these tools across 2,000+ person organisations and seen substantial failure-cost reductions. They are the right tools for products whose defects are visible.

A battery cell developing lithium plating from a subtle coating inconsistency will pass every test you run on it at the end of the line. Your control plan must prevent the process condition that creates that defect — not detect the defective cell. Detection assumes the defect is catchable. Prevention assumes it isn't, and engineers the process so the condition cannot occur.

This means process capability studies on electrode coating with the rigour we apply to turbine disc machining. Full parameter traceability on every cell — not batch averages, not sampling plans, but individual cell-level records sufficient to reconstruct production conditions eighteen months after shipment. Under AS9100, that level of traceability is baseline. In battery plants, it's aspirational.

Physics will force the convergence

Battery quality standards will converge toward aerospace logic. Not because regulators chose it — because thermodynamics demands it. The plants that survive the next decade will be the ones that separated yield from safety assurance and built their organisations around the second one.

The industry is scaling at a velocity that would be impressive if the quality architecture underneath it matched the failure physics of the product. It doesn't yet. Headlines celebrate higher standards while plants run yield dashboards borrowed from steering rack production and call it quality management. Physics doesn't negotiate, and it doesn't read press releases. The first cell manufacturer that treats defect management as latent-event prevention — not yield optimisation dressed in safety language — won't just have cleaner audits. They'll be the one still operating when the others are explaining themselves to investigators.