The concession expires. The expectation doesn't.

The leaked Foxconn memo – a customer quietly relaxing visual limits on a new folding phone's screens so a launch could survive a display shortage – is being reported as betrayal. Sacrificed standards, corners cut at the altar of the shipping date. Here is an uncomfortable observation from the other side of the table: every quality director in automotive and aerospace keeps a drawer full of the same paperwork. I could fill a binder with concessions I have signed, granted, argued over and unwound. The relaxed screen limit is not a scandal. It is a Tuesday. The real question is not whether you bend the specification in a crisis – everyone does – but whether the bend carried an exit.

What a concession actually is – a loan against future discipline

Strip the outrage away and a concession is simple: a documented, time-boxed agreement to accept product outside specified limits, signed by someone with authority on the customer's side. IATF 16949 calls it a deviation permit. Aerospace calls it a use-as-is disposition. The shop floor calls it what it is – a loan. You are borrowing throughput, or launch timing, or a quarter of capacity, and the repayment is future discipline: containment, sorting, requalification, restoration of the original limit. Like any loan, it is only cheap when the terms are written before the money moves.

When I built the QA/QC function from zero at SNOP's greenfield plant – 900+ employees, no legacy habits, no legacy excuses – the concession register had one non-negotiable rule before it had a template: every deviation carries an expiry date and a named owner. Not a department. A person, with a phone number and a signature.

A concession that never expires is not tolerance – it is amnesia with a signature.

The three clauses that separate a deviation from drift

Three clauses fit on one page, and their absence costs millions. The first is a sunset date, calendar-bound rather than event-bound, because "until the shortage ends" is not an expiry. Shortages end gradually; a limit that outlives its reason quietly becomes the weather. The second is a named containment owner – a person who personally owns the sorter, the extra inspection station, the traceability split. Ownerless containment is a cost centre that dies the day its owner changes jobs. The third, and the one most often missing, is a requalification trigger: the defined event that forces return to the original limit. Lot cut-off, tooling revision, re-run capability, re-PPAP. Without a trigger, "temporary" has no mechanism for ending.

I learned the trigger clause the expensive way. At WITTE Automotive, running the failure-cost reduction work – QRQC on the daily defects, A3 for the structural ones, a Q-Wall that made status impossible to hide – we held customer-granted concessions for limits relaxed during a capacity crunch. Granted in writing, by the customer's own engineers, to keep their line moving. Two years later their auditor opened the drawing, compared it to shipped parts and wrote findings against product sitting comfortably inside the concession. The sunset dates existed on file. What never existed was the event that would force the return, so nothing forced it until the audit did. The unwind – PFMEA rebuilt, capability re-run, samples re-submitted, stock sorted across three warehouses – taught me more about failure cost than the original crisis ever did.

The unwind costs more than the crisis

The crisis is priced. Line-down cost, expedited freight, launch delay – you can quote those in a meeting and decide. The unwind is not priced; it is discovered, usually during someone else's audit. The mechanics are unforgiving, because suppliers optimise to whatever limit you actually enforce. Within weeks of a relaxed limit the process is centred on it: tooling cut to it, control charts re-baselined to it, operators judging good parts by it. The concession expired in the register. The process kept running to it. Re-tightening means re-centring a process that has spent months practising being worse.

Who pays for that? Whoever holds the weaker paperwork. Contractually the customer granted it; commercially the supplier absorbs it unless the concession file is airtight. My cleanest quarter at SNOP – zero critical customer escalations – was not a quarter without deviations. It was a quarter in which none of them aged past expiry, because the register made ageing visible and renewal demanded a fresh signature with a fresh rationale.

The open-concession audit you should run this week

Run it before someone runs it for you. It takes an afternoon if the register is honest, and a quarter if it is not.

• Pull every expiry date older than today with no closure evidence. That count is your exposure.
• Cross-check each "temporary" limit against control plans, CMM programmes and PFMEA revisions. A relaxed limit that migrated into a controlled document is an engineering change nobody approved.
• Flag anything renewed three or more times. At that point the exception is the standard; treat it as a permanent change and process it as one.
• Name the owner of every live concession out loud. If you hesitate, you have your answer.

The Foxconn memo leaked; that is the only unusual thing about it. Most concessions never make the news. They make archives, audit findings and quiet write-offs. Somewhere a supplier will re-centre a process to whichever answer arrives, and somewhere a commercial team will discover the cost of the question. The crisis had a price that fit in a spreadsheet. The drift has a price you meet later – in a finding, in a sorting campaign, in a claim negotiation where the relaxed standard is suddenly treated as the original spec. That asymmetry is the part worth teaching: the customer grants the concession, but the expectation – the drawing, the original intent, the standard everyone remembers when it suits them – never signs anything. The concession expires. The expectation doesn't.