The certification window closes at design freeze, not at the pilot build
The certification window closes at design freeze, not at the pilot build. By the time the first pilot unit comes off the line, the three things certification actually consumes — finished samples built on the production process, a reserved slot in the test lab's queue, and a BOM revision nobody will touch afterward — have either been committed or they have not. If they have not, the pilot build can still be moved and the schedule can still be compressed, but the test window is gone, because the samples were never allocated and the queue was filled months before those pilot units existed.
I want to walk you through the freeze review, because that hour is where this gets decided.
What is actually on the table at a freeze review
You are in a room with the released drawing package, the BOM printout, and a schedule bar that says DESIGN FREEZE with a date on it. The pack is on a cart in the hallway, still smelling faintly of flux and tape, waiting to go back to the bench. Somebody from engineering is walking the room through open items — a connector that just moved, a busbar tolerance, a firmware flag that may or may not hold.
Everyone in that room is treating the date as an engineering milestone. It is not. It is a procurement milestone.
Three purchase decisions become real the moment that date is signed, and only one of them sits on the engineering side.
First, the samples. Certification testing needs finished units. Not cells, not bare boards, not a breadboard pack running off a bench supply. Finished packs, assembled the way the production line will assemble them, because a unit built on a prototype fixture is not representative of the product you are certifying. So the sample quantity has to be written into the pilot build plan as its own line item with its own serial range, not pulled off the tail of the run after everyone else has taken what they need.
Second, the queue. Lab capacity gets booked long before the samples exist. What drives the calendar is not how long the tests take, it is where you land in the queue — and that queue is full of other programs that booked before you did. The booking date is a program milestone. If the freeze slips, you are not moving your test date by the amount the freeze slipped. You are moving it to whatever opening is left, which is a different and usually larger number.
Third, the revision. Somewhere in the release paperwork is a revision letter on the BOM. That letter gets printed on the test request and it comes back on the certificate. The moment anyone runs an ECO against the cell, the protection circuitry, or the enclosure, the letter changes and the certificate no longer describes the product you are shipping.
Samples you cannot scavenge
I keep saying allocate rather than collect because several of the required tests are destructive. Whatever goes into them does not come back. If you planned to draw your test units from pilot yield, you have tied your certification schedule to the one number in the whole program you control least: how many good units the first run produces.
Plan it the other direction. Work the sample quantity backward from the test matrix, add the units your customer will want for their own verification, and put that number into the build plan as committed output — built on the same equipment and process the mass production units will use. Then pilot yield stays a yield problem, which you can work, instead of becoming a certification problem, which you cannot.
One honest caveat: on some programs the freeze is staged rather than single. Mechanical and cosmetic freeze can land weeks ahead of electrical freeze, and firmware can stay open longer than either. When that happens you can sequence tests against the staged dates instead of stacking all of them against one. Ask your supplier which of their internal freeze dates feeds which test. The answer is rarely on the schedule bar, and it is worth the twenty minutes.
Certification runs alongside PPAP, not after it
The other sequencing error I see is treating certification as a step that follows the PPAP package. They run in parallel. On a lot of customer programs the certificate is itself a required element of the PPAP submission — a line in the submission index, carrying the revision letter that has to match the drawing you are submitting against.
That puts a lab queue inside your production part approval timeline, which is a dependency the APQP plan usually does not model. If the certificate is not back when the submission goes in, you are not submitting an incomplete PPAP so much as a PPAP with a gap the customer has to disposition. Some will accept a promise with a date. Some will not. Either way it is a conversation you want to have before the freeze, not after.
And one thing needs to be said plainly: certification does not replace PPAP or APQP discipline. The certificate is one document out of many. Dimensionals, capability studies, measurement system work, the control plan, the process flow — none of that goes away because a lab signed a report. The certificate tells a customer the product is safe and compliant. It says nothing about whether your line can make it the same way twice.
The certificate carries a revision, and revisions decay it
This is the part that bites people six months later. A certificate is not a permanent property of the product. It is a document tied to one configuration, and in practice it expires the moment the configuration changes.
A cell swap. A second source for the protection IC. A different enclosure resin because the original had a lead time problem. Any of those can put you back in the queue — with a design that is already in production, and therefore with no pilot build left to carry the samples you would need.
That is why the pilot-to-mass-production transition checklist needs a line that says, in plain language, that the certificate on file matches the released BOM revision. Not certification complete. Matches. Somebody has to physically compare the revision letter on the certificate against the revision letter on the released BOM and sign it. I have watched that line catch a mismatch that would otherwise have shipped.
What I would put on the pre-freeze list
Four items, and all four are decisions rather than tasks.
A committed sample quantity in the pilot build plan, with its own serial range and its own build traveler, built on production tooling.
A booked slot in the test queue, with the booking date recorded as a program milestone on the same chart as the freeze.
A named owner for the revision letter, plus a rule that any ECO touching cell, protection circuit, or enclosure routes back through that owner before release.
A note in the PPAP index showing where the certificate lands in the submission, and which person at the customer has agreed to receive it.
Get those four before the freeze signature and the pilot build stops being your deadline. Miss them and you will spend the pilot build watching a queue you cannot enter, explaining a schedule that was never yours to keep.
