Your Approved Battery Pack Sample Is Not What Production Will Ship
An approved battery pack sample is not evidence of what production will ship; it is evidence of what one build, on one line, with one cell lot, was capable of. The approval becomes predictive only when a frozen reference sample, a written change-notice process, and production end-of-line data tied to the same measurements and limits sit behind it. Without those records, the approval is a snapshot of a sample build, and mass production is a different build.
Put both packs on the ESD mat in front of you. That is what I do at the receiving bench: the approved sample in its labeled anti-static bag, the first production unit in a plain box, and a printout of the sample's end-of-line record. I hand you the calipers and ask you to compare the enclosure seam, the label, the connector keying, and the firmware sticker. The sample has a handwritten traveler with a build date and three sets of initials. The production pack has a printed QR label. From arm's length they look like the same part. Up close, they are two different claims.
That is the moment the question stops being paperwork. The sample did not lie. It answered the question it was built to answer: can this design work when the right people are watching? Mass production asks a different question: will this line, this fixture, this cell lot, and this test station repeat it when nobody is watching?
The Sample Build Is Not the Production Process
On a sample build, the line is often a bench. The fixture may be machined for one unit or a handful. The cell lot may be a purchased sample lot that will never be used again. Tab welding may be manual or semi-automatic, with an engineer standing nearby and checking the weld appearance. End-of-line testing may be full functional testing on every unit, with the test engineer watching the screen and writing down anything that looks odd.
None of that is fraud. It is just not mass production.
The production line may use a different fixture, a new cell lot, automated tab welding, and a sampled end-of-line test sequence that runs faster than the sample's test. The operator may be different. The test station may be different. The label printer may be different. I have watched a production pack fail a dimensional check that the sample passed easily, not because the design changed but because the production fixture located the pack from a different datum. That is the kind of gap a sample approval cannot show you.
So I do not ask, "Did the sample pass?" I ask, "Which process made the sample, and where is the record that says the production process is the same one?"
A Golden Sample Is a Control, Not a Trophy
A reference sample, sometimes called a golden sample, is useful only if it stays frozen. It should carry the approved configuration: the cell part number and lot, the BMS firmware version, the enclosure revision, the harness routing, the label artwork, the build date, and the end-of-line test record that went with it. If it is sealed in a bag with a signed label and a photo of that label, it becomes a physical reference. If it is sitting on a shelf with no traveler, it becomes a memory.
I want the reference sample logged in and out. I want to know when it left the case, who compared against it, and what they found. A golden sample that gets handled every week without a log is not a control. It is a display piece.
The sample also has a limit. It can show you that a production unit is physically different from the approved configuration. It cannot prove that the production process is capable. That is what the process records are for. The sample tells you what right looked like. The records tell you whether right is still being built.
The Change Notice Is the Real Approval
A change to the cell, the BMS firmware, or the enclosure after sample approval resets the qualification evidence. That sentence is easy to write and hard to live by. The cell change is the obvious one. The BMS firmware change is the quiet one, because the pack can look identical while the behavior changes. The enclosure change is the one that shows up late, when a new tool has a different draft angle or a different connector cutout.
A written change-notice process is what makes those changes visible instead of silent. I want a change request with a number, a description of what changed, the reason, the affected part numbers and firmware versions, the disposition, and the re-qualification evidence. I want the approved sample record and the first-article record to point to the same configuration after the change is closed.
An email that says the new cell is equivalent is not a change notice. A verbal assurance at a factory visit is not a change notice. A revised drawing without a revision block is not a change notice. The sample approval is only as strong as the change-control process behind it. If changes can happen without a record, the approval is already stale; you just do not know it yet.
First Article: Follow the References
A first-article or PPAP-style submission for pack assembly is not a document dump. It is a chain of references. It normally includes dimensional results, functional results, a process flow, a control plan, and the end-of-line test record. The point is not that all those documents exist. The point is that they all reference the approved sample configuration.
I check the revision block, the serial number, the firmware version, the fixture ID, and the date. I want the process flow to name the same tab-welding process and the same test steps that the production line will use. I want the control plan to show where the end-of-line test sits and what happens when a unit fails. If the first-article end-of-line record does not reference the approved sample, I have two documents, not one chain.
A small pilot run may not justify a full PPAP binder. That is a real tradeoff. But even a low-volume build needs a frozen reference, a written change log, and an end-of-line record that shares the sample's test steps. Without those three, the first article is a ceremony.
End-of-Line Numbers Have to Match the Same Ruler
Production end-of-line test data should use the same measurements and the same limits as the sample's test record. If the sample record shows a communication check, a leakage check, and a functional sequence at final test, the production record needs to show those same steps. If production only records a quick pass or fail, the two sets of numbers cannot be compared at all.
This is not about demanding more testing. It is about demanding comparable testing. A production line can sample units instead of testing every unit, and that can be a legitimate choice. But the units that are tested need to be tested against the same limits, with the same equipment, and with the same pass or fail criteria as the approved sample. Otherwise the production data looks clean and tells you nothing.
I have seen a sample record with a detailed test table and a production record with a single green check mark. The production team was proud of the yield. I could not tell whether the yield meant anything.
What I Check Before Release
I pull the reference sample first. I confirm the seal, the label, the firmware version, the cell lot, and the test record. If the sample has been opened or reworked, I want to know why and what record was updated.
I pull the change log next. Every change after approval should have a notice, a disposition, and evidence that the affected qualification records were updated. If the change log is empty, I ask how that is possible. If the change log is full but the first-article records still point to the old configuration, I stop.
Then I pull the first-article package and the production end-of-line data. I compare the test steps, not just the pass rates. I look for the same measurements, the same limits, the same fixture, and the same firmware version. I walk the line if I can. I want to see the tab-welding station, the test station, and the label printer with my own eyes, because a record can be complete and still describe a line that no longer exists.
Finally, I keep the approved sample where it can be compared physically. Not in a drawer with no label. Not in a display case. On the bench, under a log, with the records that make it mean something. The pack that was approved is not the pack that will ship. The records are what close the gap, and only if they are strong enough to show you the gap before mass production starts.
