How a Supplier Quality Auditor Reads an IP Test Report Before Accepting It
An IP test report only earns a place in a supplier qualification file when three things line up: the specimen traces to a specific build and configuration, the test conditions match what your drawing and purchase specification actually require, and the laboratory that issued the report was accredited for that exact method. Miss any one of the three and the pass line on the certificate is decoration. What no report can establish on its own is that the production line builds what the lab tested. That part has to be proven separately, on the floor, with end-of-line leak data, reject trends, and a teardown of a unit that already passed.
I read these reports in the same order every time, and it is never the order they come printed in. You get the PDF on a Tuesday afternoon: thirty pages, a cover page stamped with a large PASS, photographs of the pack sitting in a tank. I skip all of it. I scroll to the block that describes what was actually tested, because that is where the document either becomes evidence or quietly stops being a test report.
Start with the specimen, not the result
Every lab words this block differently. Test sample. Description of test item. Specimen identification. What matters is whether it answers four questions.
How many pieces, and which ones. One is normal for immersion testing. If the report says three, find out whether all three passed or whether you are looking at the best of three.
What it was. Model number, part number, and ideally a serial number or work order you can trace back through the supplier's build records. A report that says "48V Li-ion pack, sample" with no traceable identifier is a description of an object, not a test of a product.
What build level. Prototype, pilot, or production. Labs rarely use the word prototype, but they will write "engineering sample, Rev A," and that is the same signal. I have opened reports where the tested unit turned out to be a bench build sealed by the engineering team with extra sealant applied by hand. Nothing dishonest about it. It also tells you nothing about the line.
What configuration was fitted at the time. This is the one people skip. Was the BMS powered or unpowered? Were the connectors mated with mating plugs, or left open? Was the pressure vent installed, and in which orientation? Were the labels applied? Every one of those changes the leak path. Field units ship with connectors mated and vents installed, so a test performed with a vent taped over and an open connector is a test of a different object than the one you are buying.
IPX8 and IPX9K are agreed numbers, not fixed ones
IEC 60529 defines the immersion method for IPX8 but deliberately leaves the depth and duration to agreement between the manufacturer and the user. There is no universal IPX8. The report has to state the depth, the duration, and — the part people forget — the water temperature and specimen temperature, since the test is meant to be run with the two within a few degrees of each other so the immersion does not create a pressure differential that would not exist in service.
So the comparison is not "does it say IPX8." The comparison is between the report's stated conditions and the conditions on your drawing or purchase spec. If your requirement is two meters for an hour and the report says one meter for thirty minutes, you have a mismatch, and it is a documentation problem regardless of whether the product would have passed the harder test. I have watched programs argue about this for a week when the fix was one email.
IPX9K is a different animal. The 9K designation comes out of the ISO 20653 and DIN 40050-9 line rather than IEC 60529, which is why a report headed "IP69K per IEC 60529" is worth a second look at the standard reference and the edition printed on it. The 9K test itself is more constrained — a defined jet pressure, a defined water temperature, a defined distance, a defined sweep through several angles — but the specimen still has choices baked in. Which face did they spray first. Was the pack energized. Were the connectors capped or mated. Was it standing free or mounted in a fixture that blocked one side.
One more thing in this section. A two-digit rating is two tests. I have read reports that test only the water digit and cite the dust digit from an entirely different document, or the reverse. If the file only contains half of an IP67 or IP68 claim, you do not have the claim yet.
The lab's name on the letterhead is not the lab's scope
Look for three things: the accreditation mark and number, the standard and designation exactly as the test was run, and the scope schedule — the annex listing what that lab is accredited to test. Accreditation is specific. A lab can be entirely competent and properly accredited for splash testing and not hold IPX9K on its schedule. The report can still be internally valid; it just carries much less weight when someone downstream asks what supports the claim.
Check the edition, not just the standard number. Check that the accreditation was current on the test date, which is not always the issue date. Check the bottom of the report for a subcontracting statement; if a portion of the work was farmed out, the named subcontracted lab and its accreditation should be traceable. And check that the equipment used has calibration dates that bracket the test, not just a calibration column filled in.
None of this is about catching anyone. It is about knowing whether you hold a third-party verification or a self-declaration with a nice logo on it. Those two things go into a qualification file in different places.
What the report structurally cannot tell you
Immersion testing destroys or at least compromises the units. You tested one, maybe three, and you cannot ship any of them. Even in a spotless report from a fully accredited lab, the document describes those units and nothing about how the ten-thousandth one will be built. That is not a gap in the lab's work. It is the shape of the evidence.
Which is why the report does not get accepted by itself. It gets accepted alongside a second folder.
The production folder
End-of-line leak or pressure-decay test records for the same part number, covering a real date range, with the leak-rate limit stated. I want to see how the tester itself is verified — a known-leak standard or a calibrated orifice checked on a schedule — because a leak tester that has drifted is a leak tester that passes everything.
Reject rate trend over the last sixty to ninety days, in ppm, not as an average for the year. A yearly average hides a step change, and a step change is usually the only interesting thing on the page.
Rework handling. Are failed units retested against the same limit, and do they go back into the same lot or into a separately identified container? Both answers can be acceptable. An unclear answer is not.
The control plan and PFMEA line for the sealing operation: the characteristic, the monitoring method, the frequency, and the reaction plan when it goes out of limits.
What I look at on the floor
Gasket shelf life and storage. Sealed bags, dates, FIFO, and whether the operator is reaching into an open bin at the end of the line. A gasket that has been sitting in humidity for a year behaves differently from one that came out of a sealed bag this morning, and the report has nothing to say about either.
Placement or dispensing. By hand, by jig, or by robot. If by hand, what tells the operator it is seated — a visual, a fixture, a vision check? A bead of sealant with no verification is an assumption with a schedule.
Fastener torque. Calibrated drivers, a torque log, and a sequence. An unevenly clamped groove will seal on three sides and weep on the fourth, and it will pass a leak test that runs for four seconds.
Connector seating. A written method, an audible or tactile confirmation, and a pull check if the interface matters. This comes up more than people expect on packs where the connector is part of the sealed boundary.
Vent installation and orientation. Which way the membrane faces relative to the spray direction, whether it is adhesive or welded, and how long it cures before the next step. Vents are the most common place where the validated design and the line quietly diverge.
The teardown
Then I ask for one unit off the production line that has already passed leak test, and I open it. Water staining at the seam. Dried mineral residue on terminals or on the BMS board. A gasket sitting proud of its groove, rolled during assembly, or flattened from compression set. Voids in a dispensed bead. A vent installed backwards. It takes an afternoon and it is the single fastest way to find out whether the process makes what the design says it makes.
Where I stop pressing
An IP rating is not a service-life prediction. Vibration, thermal cycling, connector mate cycles, and corrosion are separate evidence streams, and I do not try to make one test report carry them.
Smaller suppliers may not hold an accredited report for every variant of a pack. A well-documented in-house test with calibrated equipment, a defined setup, and a plan for third-party verification can be acceptable depending on the program's risk and the customer's requirements. That is a judgment call, not a rule, and I would rather make it explicitly than pretend the document is something it is not.
And I do not decide what the report obligates anyone to do commercially. That is a question for the contract people and their counsel, not for me. My scope is narrower: whether the evidence supports the claim, and whether the process supports the evidence.
The report is a claim about one or two units. The audit is a claim about a process. I keep them in that order, and I do not let the first one stand in for the second.
