PPAP Readiness Score: Why "Complete" and "Ready to Submit" Are Two Different Numbers
A supplier quality engineer pulls up a Level 3 PPAP submission at 9 p.m. the night before it ships. All eighteen elements show a green checkmark. The spreadsheet tracker says 100 percent. She sends it anyway, because the tracker has never once caught a real gap, it only counts whether a document got uploaded to a folder. Nine days later the customer reviewer kicks it back: the control plan lists a bore diameter as a Critical Characteristic, and there is no capability study anywhere in the package for that dimension. The element for Initial Process Studies was there, it just covered the wrong characteristics. The checkbox was true. The package was not ready.
That gap between "every box is checked" and "this will survive a reviewer's first pass" is the whole reason we do not report PPAP readiness as one number. We report two, we never average them, and we show a proactive list of what usually goes wrong before the reviewer finds it. Here is how that actually works.
The number a quality lead means by "ready"
When someone asks "is this part ready," they mean one thing: how many of the required elements for this submission level are actually approved. That is the headline dial on the submission's readiness panel, elements approved or waived over elements required, and it is driven entirely by element status, not by anything happening underneath in the document graph. A Level 1 submission only requires the Part Submission Warrant. A Level 3 requires fourteen of the eighteen elements. A Level 5 requires all eighteen. The dial reads against whichever set the submission actually needs, so a Level 1 part and a Level 3 part are never compared on the same scale by accident.
This headline number is intentionally simple, and it is intentionally the only thing a quality lead has to read to answer "am I ready." But a submission can hit 100 percent on this dial the same way the spreadsheet in the story above did: every element has a document attached, every box is checked, and the content behind the checkbox is still hollow. That is the exact failure mode a second, quieter system exists to catch.
What sits behind the headline, on purpose
Every characteristic, process step, and failure mode in a submission lives on a linkage graph, not just inside a document. A Critical or Significant Characteristic has to carry two specific connections: it needs a control plan line governing it, and it needs a capability study behind it. A characteristic on a Level 2 or Level 3 submission, where Dimensional Results is a required element, also needs a measured result recorded against it. A process flow step needs a failure mode analyzed against it in the PFMEA. A PFMEA failure mode needs a mitigation on the control plan. These are the same cross-document rules a human reviewer runs by hand, tracing part number and revision across Design Records, Process Flow, PFMEA, Control Plan, MSA, and Initial Process Studies, except the platform runs all of them the moment the graph changes instead of once, right before submission, under deadline pressure.
We call this dial "Structure" on purpose, and we keep it one click away behind a "Details" disclosure rather than on the headline, because a graph can look wired up and still be empty of judgment. An edge existing between a characteristic and a capability study proves a study got attached. It does not prove the study reports the right index, or that the control plan line governing that characteristic is the one that actually controls it on the floor. Naming this dial "Structure" and keeping the headline reserved for element completion means nobody on the team can read "100 percent complete" off an empty package whose graph edges merely happen to be present. That distinction is the product decision, not an afterthought bolted on after a bug report.
Structure quality: the number that catches what a checkbox and a wired edge both miss
Sitting next to Structure is a second score out of 100, Structure quality, built from two dimensions that are scored independently and then combined, never folded into the completeness percentage above them.
Capability soundness asks two questions on every linked capability study: does it report a Ppk or Cpk value at all, and does that value clear the minimum index for the submission level, 1.33 for a Level 2, 3, or 5 submission where Initial Process Studies is required. Each question is worth points on its own, so a study that reports a number below threshold scores differently from a study that reports no number at all. On top of the deterministic check, an AI pass reviews the same evidence for the thing a rule cannot see, whether an initial study correctly reports Ppk, the long-term index appropriate to a first production run, rather than Cpk, and adjusts the score by as much as five points in either direction with a one-sentence rationale attached. We cover the Ppk-versus-Cpk distinction and the three-tier acceptance criteria in full elsewhere, but the short version is that a beautiful Cpk number on a brand-new process is not evidence of anything an initial study is supposed to prove.
Traceability integrity asks whether the graph itself holds together where it matters most: no Critical or Significant Characteristic missing its control or its capability link, no required cross-document edge missing anywhere across the characteristic spine. The AI layer here judges something structural checks cannot, whether the control governing a characteristic is actually the right control for that characteristic, not just any control plan line that happens to be linked.
Both dimensions are worth fifty points each. Neither is averaged with the Structure percentage above it, and the two dimensions are never averaged with each other either, on purpose. A quality lead who sees Structure quality at 45 and clicks in can tell immediately whether the weak dimension is capability evidence or traceability, and that tells them exactly what kind of gap to close, not just that a gap exists somewhere.
Where this catches the bore diameter before day nine
Back to the Level 3 submission from the opening. If that package had gone through this evaluation before it shipped, the story plays out differently. The bore diameter, marked Critical Characteristic on the drawing, has a control plan line but no linked capability study. Structure records a missing required edge, severity critical, the highest tier the system assigns, ranked above a missing edge on a plain, non-special characteristic. Traceability integrity loses its full allocation for the missing-critical-link check. And separately, before anyone even opens the readiness report, a proactive "consider" finding surfaces pulled from a library of common non-conformances tied to the four elements most submissions live or die on: Design Records, Control Plan, Dimensional Results, and Initial Process Studies. That finding names the exact pattern this package is walking into, a capability gap on a designated characteristic, before the customer reviewer ever sees the package.
The element checkbox for Initial Process Studies can still say submitted, because a study exists in the package for some characteristic. It just is not this one. That is precisely the distinction "Complete" and "ready to submit" are built to keep visible instead of collapsing into a single reassuring percentage.
The other direction: fully wired, still not sound
The opposite failure is just as real and just as expensive. A package can be structurally perfect, every required edge present, Structure reading 100 percent, and still carry a capability study that reports Cpk instead of Ppk on an initial study, or a control plan line linked to a characteristic it does not actually govern in practice. Nothing about the presence of an edge catches either problem. That is why capability soundness and traceability integrity exist as AI-reviewed dimensions sitting on top of the deterministic graph check, not as a restatement of it. The same design and process seam shows up whenever a DFMEA hands a critical characteristic down to a PFMEA and control plan: the link can exist and still point at the wrong control.
Reading the Structure quality number
Structure quality is not a pass or fail gate on its own, it is a graded read that tells a quality lead what kind of week is ahead. A score in the high eighties or above means the package content is sound and the remaining work is procedural. A score in the sixties or seventies means the package is solid but one dimension, usually named directly in the summary copy, needs tightening before submission. A score in the forties or fifties means there are several real content gaps to close, and a score below forty means the package needs substantive work, not a final polish pass, before it is worth a reviewer's time at all. The point of grading it this way instead of a single threshold is that a 92 and a 58 both deserve attention, but they do not deserve the same Tuesday.
Three questions this comes up in practice
Can a package be 100 percent Structure and still fail Structure quality? Yes, and that is the case that matters most. Every required edge can be present, every characteristic linked to a control and a capability study, and the AI layer can still catch that one of those capability studies reports the wrong index or that a control plan line does not actually govern the characteristic it is linked to. Structure proves the graph is wired. Structure quality judges whether the content behind the wiring is defensible.
Does a low Structure quality score block submission? No. The score is a readiness signal for the team, not a workflow gate. A part can still move through approval, revision request, or rejection regardless of what either dial reads. The dials exist so the decision gets made with the real picture in front of the reviewer instead of after the customer finds the gap.
Why does the AI adjustment only move a dimension by five points in either direction? The deterministic rule checks carry the majority of each dimension's score on purpose, because whether a required edge exists or a Ppk value clears 1.33 are facts, not judgment calls. The AI layer is there for the narrower question the rules cannot answer on their own, whether the evidence behind a passing edge is actually sound, and a five point ceiling keeps that judgment call from ever overriding what the deterministic graph already proved or disproved.
The standards spine underneath this
IATF 16949 Clause 8.3.4.4, Product Approval Process, requires the organization to run a product and manufacturing approval process that conforms to what the customer defines, prior to submitting a part for approval. The clause does not name PPAP directly, it leaves the specific approval process to the customer, but the AIAG Production Part Approval Process is the method nearly every automotive customer specifies to satisfy it. That is why the elements behind this scoring model map directly to AIAG's eighteen: Design Records is Element 1, Control Plan is Element 7, Dimensional Results is Element 9, Initial Process Studies is Element 11. Those four are exactly the elements the "consider" findings watch first, because the completeness of the 18-element package as a whole still comes down to whether the highest-risk elements individually hold up, not whether every folder has a file in it. Measurement System Analysis, Element 8, sits one step upstream of capability soundness for the same reason, a capability study is only as trustworthy as the gauge that produced the data feeding it.
What this takes off a quality engineer's night
The manual version of this review is a person opening Design Records, Process Flow, PFMEA, Control Plan, MSA, and Initial Process Studies in six different tabs, cross-referencing part numbers and characteristic classifications by eye, and trusting their own memory to catch the one link that got missed on a Tuesday three weeks ago. It is exactly the kind of review that goes fast under deadline pressure and misses the thing that matters most, because a Critical Characteristic and a cosmetic one look identical in a stack of PDFs.
Because Build generates the process flow, PFMEA, control plan, and inspection plan from the same underlying graph, and Package reads that graph directly, the Structure and Structure quality dials update the moment a control plan line changes, not the next time someone remembers to re-check it by hand. A capability study run in Analyze links straight back onto the characteristic it belongs to, closing the exact edge that catches the bore diameter problem above. None of this replaces the reviewer's judgment. It puts the two numbers that judgment actually depends on in front of the engineer before the package leaves the building, not nine days after a customer reviewer finds them instead.
QualityEngineer.ai runs a 30 day trial with no credit card required if you want to see what your current PPAP packages score on both dials.
