Three headlines landed in the same week. Vertical Aerospace becomes the first eVTOL developer to hold AS9100. Rolls-Royce onboards a new civil-parts maker in India. The trade press runs features on scaling aerospace production without costing quality. Read them together and the pattern is obvious: the entire sector is stepping rate at once. Here is an uncomfortable observation from two decades of doing this — the most dangerous week in a quality career is the one where output doubles and the PFMEA stays on the shelf.
I have lived that week. I built the greenfield QA/QC department for a 900+ employee automotive plant and held a 70% defect-cost reduction while volume scaled, which is a very different discipline from holding quality on flat production. Nobody throws a party for the rate increase. The banners go up, the takt board turns over, and the failure mode and effects analysis — written, validated and signed off at yesterday's rate — quietly becomes a description of a factory that no longer exists.
Rate is a process variable, not a business number
Business people treat rate as a commercial decision. On the shop floor it is a process variable with the same standing as temperature, torque and feed speed. Change it and you have changed the process, whether or not anyone filed the paperwork.
Doubling takt breaks things in specific, predictable ways. Tool life assumptions collapse first: a cutter rated for 30,000 cycles at eight-second cadence behaves differently when the spindle never cools between parts. Thermal drift compounds. Machines that stabilised by mid-shift now run hot all shift, and last year's capability study was done on a cold machine at half the load. Operator cadence degrades in ways no training matrix captures — the inspection that took ninety seconds when three parts an hour crossed the bench gets compressed to forty-five when six do. The inspector doesn't decide to skip it. The takt decides for them.
Worst of all, the occurrence rankings in your PFMEA are empirical measurements of a process that ran at yesterday's rate. Double the rate and every "occurrence: 2" on that sheet is now an unverified guess. Under IATF 16949 and AS9100 I have watched teams defend rankings that were demonstrably measured under conditions that no longer applied. The document was correct once. Correct once is not correct.
A rate increase is an engineering change that forgot to file itself.
Where ramps actually break
Ramps fail in three places, and almost never where management is looking.
Inspection capacity becomes the bottleneck, so sampling quietly gets cut. Nobody approves the reduction. The control plan says check every fifth part; the takt now delivers parts faster than the check can be performed, and the operator adapts — to every ninth, every twelfth. The control plan is untouched. The reality has moved. I have walked lines during ramp weeks where the documented sampling frequency and the actual one had diverged so far the paperwork was closer to fiction than description. We caught it with QRQC, pulling the team onto the floor within the hour — not through an audit three months later.
Suppliers capable at pilot volume are unproven at serial volume. A supplier who shipped you 400 good parts a month can be a different company at 4,000. Their tooling, their incoming inspection, their operator training — all calibrated to a volume you just invalidated. At volume thresholds a re-PPAP is not bureaucracy. It is the only honest answer to the question "does their process still produce what we qualified?"
New-hire density dilutes tacit knowledge. When I led quality initiatives across a 2,000+ workforce on multiple sites, ramp hiring meant some cells ran with a majority of operators under six months of tenure. The veteran who used to hear a bad bearing before the gauge caught it is now one supervisor covering three shifts of strangers. Escapes that used to die at the bench walk through. That is not a people failure. That is a control plan that assumed a staffing profile it no longer has.
Treat every rate step as an engineering change
The fix is boring and it works. A rate step goes through change control like any other process change, with rate gates:
- Re-run capability at each rate gate. Not a review of last year's Cpk — a fresh study at the new takt, on warm machines, with current tooling.
- Refresh occurrence rankings with rate-adjusted data. If failure frequency scales with cycles, the rankings must scale too, and the PFMEA revision must precede the rate increase — not trail it by two quarters.
- Re-balance control-plan sampling to takt. If inspection cannot keep pace, add inspectors or change the method. Do not let the floor improvise the answer.
- Trigger supplier re-PPAP at volume thresholds. Define the threshold contractually before the ramp, so it is a plan, not an argument.
Run the honest cost comparison. A few thousand euros of requalification — capability studies, revised PFMEA hours, a re-PPAP — against a seven-figure escape, customer containment and, in aerospace, regulatory exposure. This is not a close call. During ramp weeks I leaned on A3 discipline precisely because the temptation to sort it out later is strongest exactly when the business is loudest about output. The A3 forces the rate change to argue its case on paper before it argues it with scrap.
For the wave of newly certified producers entering serial aerospace production — the first AS9100 eVTOL developer, the new civil-parts makers being onboarded in India and elsewhere — this matters twice over. Your certificate was earned at pilot rate. The paperwork that got you certified describes a factory about to triple its takt. The certification body will not chase you on this. Physics will.
Key takeaways
- Treat any rate change above roughly 20% as a formal engineering change: PFMEA revision, control-plan review and fresh capability study before the new takt runs.
- Occurrence rankings are empirical claims about a specific rate — revalidate them whenever takt changes, because stale rankings fail silently until they fail publicly.
- Audit sampling frequency against actual takt during ramps; unapproved sampling cuts are the most common invisible nonconformity on a stepping line.
- Set contractual supplier re-PPAP volume thresholds in advance, so requalification at scale is scheduled work, not emergency work.
The question I would put to any plant manager this week is simple. Walk onto your floor, pull any PFMEA from the rack, and ask whether a single page of it still describes the takt running in front of you. If the honest answer is no, you are not managing a ramp. You are managing an undocumented process change that has already been implemented — by the production schedule, without your signature. The rate doubled. The paperwork didn't. Somewhere in that gap, an escape is waiting for its moment.