The yield didn't collapse. It was scheduled.

"Low yields and quality concerns." The reports around this autumn's flagship phone read like a discovery – supplier sources, production hiccups, weeks left on the clock. I have lived enough launches, automotive and aerospace, to tell you what that phrase is. Not news. A number that sat in a CpK report months ago, seen by everyone with a calendar invite and priced in.

Two decades of launches and I have never seen a yield problem surface in the final weeks. Acknowledged in the final weeks, yes. Different event entirely. By then the plant has lived with it since the first run-at-rate, finance carries a line for it, and the argument about escalating was lost in a conference room months earlier.

Yield loss on a fixed date is a scheduling decision

Every launch calendar I have worked to was drawn backwards from a date nobody in the plant was allowed to move. Automotive calls it J1 or SOP; consumer electronics calls it a keynote. The date exists first, and the ramp curve is fitted to it afterwards. Once it cannot move, there are exactly three ways to hit it: be capable on time, buy capacity you do not need, or convert capability risk into scrap. Only one of those requires no meeting.

When a process runs at a CpK of 0.9 against a spec that demanded 1.33, the yield is not a mystery. It is arithmetic. On a 50,000-unit ramp with €40 of material per unit, every point of yield is worth €20,000 – and the capability study forecast the total months before first shipment. The failing units do not vanish. They get bought twice, sorted, reworked or written off under a heading like "ramp attrition". Nobody calls it scrap, because scrap sounds like failure. A ramp curve sounds like progress.

A ramp curve drawn backwards from a marketing date is a scrap forecast with a better name.

The uncomfortable part is the consent. The supplier sees the number in its own capability study, the programme team in yield tracking, finance as launch-period margin compression. Everyone assumes someone else has escalated. Nobody has, because escalating after the gates passed would mean admitting the gates were waved through.

Capability before rate – the gates exist for this

APQP exists so capability is proven before volume, not discovered by it. PPAP, run-at-rate, capability studies on significant and critical characteristics – this is not paperwork ritual; it is the mechanism that keeps unknown capability from becoming planned scrap. Six Sigma is the same idea with statistics instead of signatures. Under launch pressure I have watched teams waive those gates in three ways, always "temporarily":

The interim approval. A limited PSW signed "until the tooling is corrected" becomes permanent, because the correction competes with next year's launch for the same capital budget.
Containment instead of capability. One hundred percent inspection replaces the capability study. Sorting becomes the process, three extra inspectors become headcount, and nobody asks why a feature the machine was supposed to hold now needs eyes on every part.
Moving the target. Yield is measured against a "maturity curve" that is the scrap forecast inverted, so the process is on plan every month by definition.

Every waiver carries the same signature: it transfers risk from the calendar to the floor.

Someone must own the words "not ready"

When I built the QA/QC function for SNOP's 900-plus employee greenfield plant – empty floor, no history, everything from zero – the safe-launch phase was where the plant was won or lost. Daily QRQC at the line through the whole ramp: defect named within 24 hours, owner and countermeasure on the board, A3 for anything chronic. The discipline is not pleasant. It is necessary, because a ramp without a daily clearing mechanism buries its defects in shipment. We finished the first quarter with zero critical customer escalations. That was not luck. That was problems dying small.

Red-button authority matters as much as cadence. One named person must be able to say "not ready" and stop the ramp without convening a committee – and must see leadership back them when they use it. A stopped line costs a day. A ramp that starts incapable costs eighteen months of degraded process: the containment that became standard work, the hand-finishing station that acquired a budget line, the PFMEA that no longer describes the line it is filed under.

And when yields dip – mine have, everyone's have – customers ask a narrower question than you expect. Not why. Three things: did you detect it, did you contain it, did you tell us before our own people found it. A customer who hears about your capability gap from your morning report will work with you through it. A customer who finds it at their receiving inspection will spend two years auditing you for it. I have de-escalated both kinds of relationship. The second costs a multiple of the first.

Key takeaways

If the launch date is fixed, treat the capability forecast as seriously as the volume forecast – the gap between those two curves is your scrap budget. One page, and someone senior signs it.
Give every waiver an expiry date, a named owner and a standing review. Interim approvals that never

FROM THE FIELD MANUAL

The Cost of Poor Quality Playbook

Make the invisible factory visible — the method that recovered €2M+ in hidden losses, step by step.

Get the playbook — $19