PPAP explained: the 18 elements, submission levels 1–5, run-at-rate, and what the submission is actually proving

PPAP — the Production Part Approval Process — is the evidence package a supplier puts in front of a customer before shipping production parts: the drawings and changes it worked to, the risk analyses, the control plan, the measurement studies, the dimensional and test results, the capability numbers, sample parts, and a one-page declaration signed by somebody who is accountable for all of it. The customer reviews the package, and either approves the part for production shipment, approves it on conditions, or sends it back.

That is the textbook description, and it is accurate, and it is why most PPAPs are bad. Read that way, PPAP is a document-collection exercise: eighteen boxes, a deadline, and a binder that grows until the boxes are ticked. Suppliers who treat it like that tend to pass the submission and fail the launch, because the binder was assembled to satisfy a reviewer rather than to answer the question the reviewer was actually asking.

A PPAP is not a pile of paperwork. It is an argument, made with evidence, that the process you are about to run every day — with production tooling, production people, production gauges and at production speed — will make parts that meet every requirement, every time, and that you will tell the customer before you change any of it. Every element exists to prove one piece of that sentence.

This is the third article in the Core Tools series. The control plan and the Process FMEA come first because PPAP is where both of them are submitted for the first time, and it is followed by measurement systems analysis, because one of the eighteen elements is the evidence that your measurements can be believed. The worked example throughout is the same welded pump mounting bracket that runs through the free Control Plan Template.

What the submission is actually proving

Strip away the element numbers and a PPAP makes four claims. It is worth holding them in your head while you build one, because each element should visibly support at least one of them — and an element that supports none of them is usually a sign it was produced for the binder.

1. We understood the requirements. The design record we built to is the current one, every engineering change is reflected, and every characteristic on the drawing and in the specifications has been identified, including the special ones.
2. We built a process that can meet them. The process flow, the PFMEA and the control plan describe a real process; the risks were analysed by people who know it; the controls are proportionate to the risks.
3. We proved it, honestly, under production conditions. The parts were made on production tooling and equipment, by production operators, at production rate. The measurements were taken with gauges whose capability is known. The results show conformance on every characteristic and capability on the important ones.
4. We will keep it that way, and tell you if we change it. The control plan will be followed, the records will be kept, and any change to the product, process, material, location or supplier will be notified before it ships.

The fourth claim is the one people forget is in there. A PPAP approval is not a pass mark for the parts in the box. It is the customer approving a specific process, and from the moment of approval that process is frozen in a contractual sense: changing it without telling them is not a paperwork lapse, it is shipping product from a process they never approved.

Where PPAP sits, and which standard asks for it

PPAP is the output of the product and process validation phase of APQP — the point where the launch stops being a plan and starts being evidence. The reference is the AIAG PPAP manual (4th edition, 2006), which is still what most customer-specific requirements cite; German-heritage customers often use the VDA's own production process and product approval (VDA Volume 2, known as PPA) instead, and the evidence overlaps heavily. Your customer-specific requirements decide which applies, which edition, and what the customer adds on top. They always win.

For IATF 16949 suppliers the requirement sits in clause 8.3.4.4. Described in plain English rather than quoted, it asks for four things: a product and manufacturing approval process that conforms to what your customer defines; the same discipline applied to the parts and materials your own suppliers send you, approved before you submit to your customer; documented customer approval before shipment where the customer requires it; and the records of that approval retained. Quote the wording from your licensed copy when you write a procedure or a finding.

If you are certified to ISO 9001 only, nothing in the standard is called PPAP. But 8.3.4 asks for verification and validation that design outputs meet requirements, and 8.6 asks that product is not released until planned arrangements are satisfactorily completed.

The 18 elements, grouped by what they prove

The manual lists eighteen elements. Listed in number order they read like an inventory; grouped by the claim they support, they read like an argument. Not every element applies to every part — appearance approval only matters for appearance items, bulk materials have their own checklist — but every element needs a decision recorded against it, even if the decision is "not applicable, because".

Claim 1 — we understood the requirements

• 1. Design records. The drawing, CAD model and referenced specifications for the part, at the change level being submitted, including the material specifications. For a customer-designed part that is their drawing; if you hold design responsibility, it is yours. The usual failure is a submission built to a drawing one revision behind the purchase order.
• 2. Authorised engineering change documents. Any change not yet incorporated in the design record but already in the product. If a change exists only as an email, it does not exist yet.
• 3. Customer engineering approval. Evidence of engineering approval where the customer requires it — typically for test results or trials outside the dimensional scope.
• 4. Design FMEA. Where you are design-responsible. For a build-to-print supplier it is the customer's document, but the special characteristics it identifies still have to arrive in your process documents.

Claim 2 — we built a process that can meet them

• 5. Process flow diagram. Every step, including receiving, storage, transport, inspection and rework loops, numbered once for all three process documents. The PFMEA article explains why the flow sets the ceiling on everything else.
• 6. Process FMEA. The team's analysis of what each step can get wrong, built from the flow — in the AIAG-VDA 7-step form for most new programmes, with Action Priority rather than RPN.
• 7. Control plan. The pre-launch plan for the PPAP run, and the production plan you will move to once the evidence supports it. The control plan article covers what a reviewer reads first: the special characteristics and the reaction plans.

Claim 3 — we proved it, honestly, under production conditions

• 8. Measurement system analysis studies. Gage R&R, bias, linearity, stability and attribute agreement studies on the gauges and test equipment named in the control plan — at minimum those used on special characteristics. This element is quietly what makes elements 9 to 11 believable, and it has its own article in this series.
• 9. Dimensional results. Every characteristic on the design record measured and reported against its specification, usually on a ballooned drawing so each result traces to a numbered feature. Report actual values, not "OK". Five parts per cavity or process stream is a common default, but the customer specifies the quantity.
• 10. Records of material and performance test results. The tests the drawing and specifications call up — material chemistry and properties, salt spray, fatigue, function — usually summarised in a design verification plan and report, with the lab that did each one identified.
• 11. Initial process studies. Capability or performance indices on the special characteristics (and anything else the customer nominates), from the production-condition run. This is where the numbers most often get massaged; see the worked example below.
• 12. Qualified laboratory documentation. Evidence that the labs that produced elements 9 and 10 are competent for those tests: scope of accreditation for an external lab (ISO/IEC 17025 or customer-accepted equivalent), and for an internal lab, the documented scope that IATF clause 7.1.5.3 expects.
• 13. Appearance approval report. For parts with appearance requirements — colour, grain, gloss — a separate approval, usually signed off by the customer against masters.

The physical evidence

• 14. Sample production parts. Parts from the production run, in the quantity the customer asks for. They are the box the customer can open and measure for themselves.
• 15. Master sample. A part retained by the supplier, signed off with the customer, kept for the life of the part as the physical reference — for training, boundary disputes and future changes.
• 16. Checking aids. Fixtures, templates and dedicated gauges used to check the part, with their own drawings, calibration and MSA, where the customer requires them to be listed or submitted.

Claim 4 — we will keep it that way

• 17. Customer-specific requirements. Records of compliance with everything the customer adds on top of the manual — specific forms, packaging approvals, labelling, capability targets, run-at-rate formats, sub-supplier approvals.
• 18. Part submission warrant (PSW). The one-page declaration that ties the rest together; covered in its own section below.

Notice how heavily the list leans on claim 3. Seven of the eighteen elements are evidence about measurement and results. A process document can be written in an afternoon; results from a real run, measured with gauges you can trust, cannot — which is why they carry the weight.

The five submission levels — what changes is what gets sent, not what must exist

The single most misunderstood thing about PPAP is the submission level. People talk about "doing a level 1 PPAP" as if level 1 meant less work. It does not. At every level, the supplier is expected to have completed all applicable elements and to hold them on file. The level decides only how much of it is sent to the customer, and where the customer reviews it.

• Level 1 — the warrant only (plus the appearance approval report for appearance items) is submitted.
• Level 2 — the warrant with product samples and a limited set of supporting data.
• Level 3 — the warrant with product samples and complete supporting data. This is the default when the customer does not specify otherwise, and it is what most new parts on most programmes see.
• Level 4 — the warrant and whatever other requirements the customer defines.
• Level 5 — the warrant with product samples and complete supporting data available for review at the supplier's manufacturing location.

So a level 1 submission from a supplier who has not done the MSA studies is not a smaller PPAP. It is a signed warrant declaring something untrue — and if the customer's auditor asks for the file at your next process audit, the gap becomes a finding against both the PPAP clause and your integrity as a supplier. Build every PPAP to level 3 internally, whatever level you are asked to submit.

Level 5 deserves a separate warning. It looks like the least effort because nothing is posted. In practice it is the most exposed level, because the reviewer is standing on your floor and can walk from the warrant to the control plan to the gauge to the operator in ten minutes.

The part submission warrant

The PSW is one page, and it is the only element that is always submitted. It identifies the part (number, name, change level, drawing date), the organisation and manufacturing site, the reason for the submission, the level requested, the part weight, and the result of the submission — dimensional, material and performance, appearance, and whether statistical process requirements were met. Then it carries the declaration, signed by an authorised person, that the samples are representative of parts made by the production process, that they meet all requirements, and that the supporting evidence is on file.

Three practical points. Normally one warrant per part number, unless the customer agrees otherwise. The signature is a personal statement, not a formality — whoever signs it is asserting that the other seventeen elements say what the warrant says they say, so the signatory should have read them. And the "reason for submission" box matters later, because it is the first thing anyone reads when they are trying to reconstruct why a part's approval history looks the way it does.

The customer then dispositions the submission: approved (the part meets requirements and may ship in production quantities), interim approval (shipment allowed for a limited time or quantity while a stated gap is closed — a capability shortfall, a pending test — with an action plan agreed), or rejected (the submission does not meet requirements; production shipments are not authorised, and a corrected submission is needed). An interim approval is a loan, not a gift. Track its expiry date the way you would a calibration due date.

Run-at-rate: proving the speed as well as the parts

The parts in a PPAP must come from a significant production run: production tooling, gauges, process, materials and operators, in the production environment. The manual's default is a run of between one and eight hours of production, with at least 300 consecutive parts, unless the customer specifies otherwise — and many customers do, usually with their own run-at-rate format that also demonstrates capacity.

The capacity question is the one that matters commercially. A supplier can make 300 perfect parts with the best setter on the best shift, slowly, and still fail to supply the volume the business was quoted on. Run-at-rate is the customer's way of checking the quote against the floor.

Worked example — the pump mounting bracket. The bracket line is planned at 240 parts per hour. The customer's peak demand is 9,000 brackets a week, and the plant runs five days, two shifts, 7.5 productive hours a shift — 75 hours a week. So the line must deliver 9,000 ÷ 75 = 120 good parts per hour at peak.

The run-at-rate lasts two hours with the normal crew, normal changeovers and normal breaks. The line produces 446 parts, of which 8 are scrapped at the weld check and 438 are good. The demonstrated rate is therefore 438 ÷ 2 = 219 good parts per hour — 91.25% of the planned 240, and 219 ÷ 120 = 1.825 times the peak requirement. Over a 75-hour week, that rate would make 219 × 75 = 16,425 good brackets. The run also satisfies the 300-consecutive-part minimum.

The honest reading of that result is not "we have 82% spare capacity". A two-hour run with engineers watching is a best case; it contains no unplanned breakdowns, no material shortages, no absent operator. What the run proves is that the process can run at rate. Whether it will depends on the week-long efficiency you have actually seen on similar lines — and if your history says that is nearer 65% than 91%, then the real weekly figure is closer to 240 × 0.65 × 75 = 11,700, which still covers 9,000, but with far less margin than the headline suggests. Write that reasoning into the capacity section of the submission rather than letting the reviewer do it for you.

Initial process studies: the numbers that get massaged

Element 11 asks for capability — more precisely, process performance — on the special characteristics, from the production-condition run. The manual's acceptance guidance for a stable process is a performance index above 1.67 as acceptable, between 1.33 and 1.67 as possibly acceptable but needing customer agreement, and below 1.33 as not acceptable without corrective action, unless the customer sets different targets. At PPAP, before long-term data exists, the index reported is usually Ppk — calculated from the overall standard deviation of the run — because there has not yet been enough time to show the stability that a Cpk claim rests on. The difference between the two indices is the subject of Cp vs Cpk vs Ppk in the SPC series.

Worked example. The bracket's mounting-hole diameter is a special characteristic: 8.20 mm +0.10 / −0.00, so the lower limit is 8.20 and the upper 8.30. From 125 CMM readings taken in 25 subgroups of five across the run, the mean is 8.248 mm and the overall standard deviation is 0.0085 mm.

• Distance to the upper limit: 8.30 − 8.248 = 0.052 mm; divided by 3 × 0.0085 = 0.0255 gives 2.04.
• Distance to the lower limit: 8.248 − 8.20 = 0.048 mm; divided by 0.0255 gives 1.88.
• Ppk is the smaller of the two: 1.88, comfortably above 1.67.

Now the honest checks a good reviewer will make, and a good supplier will have made first. Was the run stable — does the control chart of those 25 subgroups show any special causes? A Ppk calculated from an unstable run describes that run and predicts nothing. Is the CMM that produced the readings itself studied, so that the 0.0085 mm is process variation rather than measurement noise? And were the readings consecutive, or were the "first parts" quietly left out because the tool was still warming up? The common ways of massaging a capability study — excluding awkward readings, measuring parts from a selected hour, running the study on a different machine — all produce an index that looks like evidence and is not. If the free Cpk Calculator gives you a number you do not believe, believe your doubt.

When PPAP has to be done again

Because approval freezes the process, a change to the process re-opens it. The manual lists the situations where the customer must be notified and a re-submission is required unless the customer waives it. In plain English they are:

• A new part or product.
• Correcting a discrepancy on a previously submitted part.
• An engineering change to the design record, specifications or materials.
• Using a different material or construction from the one approved.
• Production from new, additional, replacement or modified tooling — perishable tools such as drills and inserts excepted.
• Refurbishing or rearranging existing tooling or equipment.
• A change to the process or method of manufacture.
• Moving production to a different plant or a different location within a plant.
• A change of source for subcontracted parts, materials or services — heat treatment and plating included.
• Restarting production after tooling has been inactive for twelve months or more.
• A change in test or inspection method, new technique included.
• For bulk materials, a new source of raw material or a change in product appearance attributes.

Many customers also require annual re-validation — a layout inspection and functional verification at the frequency the control plan sets, which is what IATF clause 8.6.2 expects — and records must be kept for the time the part is active plus one calendar year, unless the customer says otherwise.

The practical problem is not knowing the list. It is that the changes that trigger it rarely announce themselves. Back to the bracket: after a customer complaint about a short weld, the plant locked the weld schedule into the robot and added automatic parameter monitoring with a cell stop. That is a change of process method at OP20, so the customer was notified, the control plan went to Rev 4 with the complaint reference on its revision line, and a level 3 update was submitted and approved. That sequence is what the header block of the free template's worked example records. The failure version of the same story is a maintenance technician swapping the weld controller for a different model on a Sunday — no engineering change note, no notification, and a supplier now shipping from an unapproved process without knowing it.

The defence is the same one-line rule the earlier articles in this series recommend for the control plan and the PFMEA: every engineering change note, maintenance change and supplier change closes with the written question "does this need customer notification or PPAP?" — yes or no, with a name against it. Add the equipment and tooling changes from your maintenance system to the trigger list; that is where most unapproved changes are born.

What gets submissions rejected

Reviewers see the same problems again and again. None of them is subtle:

1. The documents disagree with each other. Operation numbers that differ between flow, PFMEA and control plan; special characteristics on the drawing that do not appear in the control plan; a control plan naming a gauge the MSA file has never heard of.
2. Dimensional results that are not results. "OK" in the result column, a missing balloon, a characteristic measured on one part instead of the quantity required, or a result reported to fewer decimal places than the tolerance needs.
3. Capability from the wrong parts. Too few readings, readings not in production sequence, a study run on prototype tooling, or a Ppk with no stability evidence behind it.
4. MSA missing or on the wrong gauge. A study on a gauge "of the same type" rather than the one in the control plan, or a %GRR quoted with no indication of what it was a percentage of.
5. The run was not a production run. Engineers operating the line, a pre-production tool, parts made before the process was finalised.
6. Sub-supplier approval missing. Your PPAP rests on your suppliers' parts and materials, and the clause expects those to be approved before you submit.

Start this week

Pick one part that is already approved and pull its PPAP file. Read the warrant, then compare it with what is running today: is the change level on the warrant the change level on the floor? Is the tooling the same tooling? Has the process moved, a supplier changed, the gauge been replaced? Then open the engineering change and maintenance records since the approval date and ask, for each entry, whether it should have triggered notification. Most plants find at least one change that did, and that one finding is worth more than any PPAP training course.

For new parts, start the file at the beginning of the launch, not the end. The process flow, PFMEA and control plan are the first three things a reviewer reads after the warrant; if they are written in the last week, it shows.

The free Control Plan Template below is the pre-launch and production control plan in the standard column set, with the bracket example's reaction plans and a 12-point health check — element 7 of the package. Its companion, the free Process Flow Template, covers element 5 with the same operation numbers. The PFMEA Toolkit completes the chain for element 6 with the AIAG-VDA 7-step form and Action Priority built in, and the clause explorer has the full IATF clause behind the requirement.

Frequently asked questions

Is PPAP the same as a first article inspection? No. A first article inspection proves that one part conforms. PPAP proves that the process making it is understood, controlled, measured and capable, at rate. A first article is roughly what element 9 alone would give you.

Does a level 1 submission mean we only need the warrant? You only need to send the warrant. You still need every applicable element completed and on file, because the warrant declares that it is.

Do we need PPAP if we are only certified to ISO 9001? Only if your customer requires it. But the questions it answers — is the design record current, is the process controlled, are the results real — are what ISO 9001 clauses 8.3.4 and 8.6 already ask, and a PPAP-style package is a tidy way to evidence them.

What does interim approval allow? Limited shipments — by time or quantity — while a specific gap is closed under an agreed action plan. It expires; shipping past its limits is shipping unapproved product.

How long do we keep PPAP records? For as long as the part is active, plus one calendar year, unless the customer specifies longer.

Interpretive guidance only — not a substitute for the AIAG PPAP and APQP reference manuals, VDA Volume 2, IATF 16949 or ISO 9001. Work from your licensed copies, and follow your customer-specific requirements, which take precedence throughout. Clause numbering follows ISO 9001:2015 and IATF 16949:2016. ISO 9001:2026 was published on 16 September 2026; the ISO 9001 clauses linked here (8.3.4, 8.6) carry over in the published text.