Pratt & Whitney Canada's decade-long forging agreement with Bharat Forge has been read as geopolitics – India rising, supply chains diversifying, another stamp on somebody's map. I read it as physics. Somewhere in Pune sits the die that will make the first articles behind this deal. That die – fresh, gauged, photographed for the qualification file – is the best version of itself it will ever be. The two-hundred-thousandth part will come from a tool that has been refurbished, re-cut and run back in, on a press whose frame has stretched a few hundredths of a millimetre since anyone last checked. Contracts and tool life run on different clocks. Two decades in automotive and aerospace quality taught me that nobody signs for the gap.
What the first article actually certified
A first article is a photograph, not a video. It certifies one moment in the life of a system: a new die, a freshly calibrated furnace, the supplier's best crew on its best shift, an incoming steel lot sitting dead centre of the chemistry window. FOREAST consulting for ArcelorMittal drummed the material layer into me – two heats inside the same specification window can forge differently, and the lot behind your first article is gone long before your second arrives. None of that appears on the certificate. The AS9102 file records dimensions, hardness, microstructure: outputs. It cannot record grain flow responding to a cavity that will wear five tenths, because the certificate is printed before the wear happens. You approved the system at its peak. Then you spent a decade shipping parts from its middle age.
Wear never files a change request
Change control is built for discontinuities – an engineering change notice, a new revision, a swapped insert. Die wear is not a discontinuity. It is erosion measured in microns per ten thousand strokes; press frames elongating under cyclic load; quench-polymer concentration drifting a point a month; thermocouples ageing until every recorded temperature reads kinder than the truth. The paperwork stays silent because nothing happened. That is the opposite failure of the stale PPAP: the process is perfectly intact, unchanged, compliant – and that is exactly the problem. Control plans sample at a fixed frequency whether the die sits at two thousand strokes or two hundred thousand. Reaction plans trigger on defects. They never trigger on drift.
VDA 6.3 process audits ask the right question – how do you manage tooling? – and get the wrong answer, because the logs capture scheduled interventions, not the re-cut a toolroom performed between shifts when maintenance ran long. I ran those audits myself, at Witte Automotive and later at SNOP, leading quality initiatives across a multi-site, 2,000-strong workforce. One forging supplier re-cut cavity two on a hot-forged component at roughly 120,000 strokes. No entry anywhere. Dimensions came back into spec at the flash line, but the re-cut moved metal differently and grain flow shifted with it. A critical dimension crept 0.09 mm over five months – inside tolerance at every monthly sample. The customer's validation engineers then found fatigue cracks the capability reports said could not exist. The tooling log was immaculate. That was our first clue. Containment, sorting and premium freight ran past €220,000, and the quarter closed with zero critical customer escalations – not because we argued well, but because we stopped auditing paperwork and started counting strokes.
A first article certifies the tool's best day; the contract buys its average decade.
Governing the decade, not the deal
The timing matters. The same week brought news of Hanwha committing $2.2 billion to a munitions campus in Arkansas and Hical opening a new aerospace plant in Bengaluru. Aerospace is locking in decade-long single-source relationships at the exact moment new capacity is being built everywhere, and the governance inside those relationships will decide more about delivered quality than any groundbreaking. None of this is a dig at Bharat Forge; the best forger on earth still runs physics. The agreement needs to govern the physics:
- Cavity-level part counters. Every cavity is a ledger with a serial number. Refurbishment gates trip at defined stroke counts, and a refurbishment means requalification of that cavity – no exceptions, no between-shift favours.
- Requalification by cumulative count, not calendar. Calendar years measure contracts. Tools measure strokes. A cavity running three shifts a day deserves its year-six review in year three.
- Streamed SPC, not quarterly decks. Plot critical characteristics against the tool-life curve, subgrouped by cavity, streamed live to the buyer. A drift that crosses nobody's tolerance can still cross a fatigue threshold.
- Resident engineering, not audit tourism. Rotate buyer engineers through the forge on a standing schedule. The year-six conversation – tooling strategy, press overhauls, capacity creep – gets booked in year zero, because the contract forgot to schedule it.
Key takeaways
- First-article qualification certifies a system at its peak; a decade-long deal consumes it across the whole life curve. Govern the curve, not the snapshot.
- Die wear, frame elongation, quench-polymer drift and thermocouple ageing are continuous changes your change control cannot see. Trigger on degradation trends, not only on defects.
- Put tool life in commercial terms inside the long-term agreement: stroke counters, refurbishment gates and count-based requalification per cavity.
- Stream SPC by cavity against the tool-life curve and station resident engineering – the buyer's real deliverable is governance, not the purchase order.
Long-term supply deals are marriage contracts written as dating rules – heavy on courtship metrics, silent on who watches the slow decay. The purchase order is not the buyer's deliverable. The deliverable is the governance that notices tenths of a millimetre while the change log stays silent, and that knows, at any hour of any year of the decade, exactly how many strokes that die has left.