APQP and new part introduction

Every problem found at launch existed months earlier. It was cheaper then, and usually visible to somebody. APQP is the attempt to make that somebody’s observation land in a defined place, at a defined time, in front of people who can act on it — which sounds like bureaucracy until the first programme that skips it.

A cross-functional project team standing around a large printed timing plan on a meeting room wall, with coloured milestone markers and sticky notes on it, first-off sample parts and a measurement report laid out on the table in front of them. — The plan on the wall is the deliverable. The parts on the table are evidence for one gate of five.

Five phases, and the one everybody drops

A diagram of the five APQP phases drawn as overlapping horizontal bands along a timeline. Phase one, plan and define, lists inputs of customer requirements and voice of the customer and outputs of design goals, reliability targets and a preliminary process flow. Phase two, product design and development, lists outputs of the design FMEA, drawings, material specifications and prototype control plan, with design freeze marked at its end. Phase three, process design and development, lists outputs of the process flow, process FMEA, pre-launch control plan, floor plan and packaging specification. Phase four, product and process validation, lists outputs of the production trial run, measurement system analysis, capability study, run at rate and the PPAP submission. Phase five, feedback and corrective action, is drawn extending into production, with arrows returning to the FMEAs and control plan of phases two and three. Gate review symbols are drawn between each phase. A note beside phase five marks it as the phase most often abandoned. — Phase five is not the end of the project; it is the arrow back into phases two and three. Drop it and every programme rediscovers the same failure modes.

Phase	Produces
1. Plan and define	Design goals, reliability targets, preliminary process flow, feasibility review
2. Product design and development	Design FMEA, drawings, prototype control plan — ends at design freeze
3. Process design and development	Process flow, process FMEA, pre-launch control plan, floor plan, packaging
4. Product and process validation	Trial run, MSA, capability study, run at rate, PPAP submission
5. Feedback and corrective action	Lessons fed back into the FMEAs, control plans and standards

Phase five is the one that quietly disappears once production starts and the team is moved to the next programme. It is also the only phase that makes the organisation better rather than the part — which is why the same failure modes keep reappearing in companies that do the first four well.

The feasibility review is the cheapest hour of the project

At quotation, a requirement can still be challenged. Afterwards it cannot, because the price is agreed and the alternative process is no longer available.

So the feasibility review asks a blunt question: can we actually make this, with the process we are proposing, to this specification? Accepting a tolerance the intended process cannot hold turns into a concession request eight months later, from a much weaker position.

Gate reviews decide on evidence

A gate review is not a progress meeting. It is a go or no-go on whether the phase’s defined outputs exist and are acceptable.

Design freeze and the float nobody counted

A timing plan diagram showing a chain of dependent tasks from design freeze through tool design, tool manufacture, tool trial, measurement and capability study, run at rate and PPAP submission to the launch date, with float drawn as a shaded buffer before launch. A second version of the same chain below shows a late design change accepted after freeze, with tool rework inserted, each subsequent task shifted right, and the float entirely consumed so that the PPAP submission now overlaps the launch date. A third version shows a task slipping without the successor dates being moved, so the plan still displays the original launch date although no chain of tasks supports it, annotated as failing quietly. Arrows indicate the only remaining recovery options as overtime, air freight or a late launch. — Late changes do not consume the schedule evenly; they consume the float at the end, which is the part that was protecting the launch.

Tooling is cut from the frozen design. A change after freeze means modifying or remaking steel, and the time comes out of whatever float remained — which is the buffer that was protecting the launch, not some general slack in the middle of the plan.

The related failure is subtler. A task slips, and the dependent dates are not moved. The plan continues to display a launch date that nothing supports, and the discrepancy only becomes visible when the last task cannot start. By then the options are overtime, air freight, or a late launch.

Run at rate — three things at once

A trial that makes good parts proves the process can. Run at rate asks whether it will: the full contracted rate, sustained, with production tooling, production operators and production material.

All three conditions matter together, because the failures that appear only under the combination are the ones that ruin launches:

Tool heating — dimensions drift once the die reaches its steady-state temperature, which never happens in a short trial.
Operator fatigue and rhythm — cycle times that are achievable for twenty parts, not for eight hundred.
Material handling and logistics — racks, containers, changeovers, and where the swarf goes.
Changeover and restart, which a continuous trial never exercises.

Three control plans, on purpose

Version	When	Character
Prototype	Phase 2	Heavy dimensional and material verification
Pre-launch	Phase 3–4	Tighter and more frequent than production
Production	After capability is demonstrated	Relaxed with evidence

Before capability is proven, the process is protected by extra inspection. Once data shows it is capable and stable, that inspection is reduced — with the data as justification. Skip the pre-launch version and you either over-inspect forever or relax the controls on nothing but optimism. The documents themselves are covered in control plans and process flow.

What actually keeps a programme on time

A cross-functional team from phase 1, including manufacturing and quality, not engineering alone.
Named owners for each output. A deliverable owned by a function is owned by nobody.
Honest replanning. Move the dependent dates when a task slips, even when the launch date then looks bad. Especially then.
Long-lead items identified early — tooling, gauges, checking fixtures. A checking fixture ordered after the parts arrive is a guaranteed delay.
Phase 5 resourced, before the team is dispersed.

What to take away

APQP is front loading: find it in planning, not at launch.
Five phases, and phase five — feedback into the FMEAs — is the one that gets dropped.
The feasibility review is the last cheap chance to challenge a requirement.
A gate review decides on evidence, and a gate passed on a promise is never revisited.
Changes after design freeze consume the float that was protecting the launch.
A plan whose successor dates are not moved reports a date nothing supports.
Run at rate combines full rate, duration and production conditions deliberately.
A hand-finished sample conceals exactly the variation being asked about.
Three control plan versions exist so controls are relaxed with evidence, not hope.
Long-lead gauges and fixtures are ordered in phase 3, not when parts arrive.