Verification is not a document collection exercise. It answers two separate questions — can this supplier make the part repeatedly, and do the papers describe the part you are buying — and the second is easier to fake than the first.
Separate capability from paperwork
Capability means the tool is right, the process is stable, and the same dimensions come out next month. Paperwork means someone can produce a certificate with your product name on it. A supplier can be strong on one and weak on the other, and the two failures look completely different in production.
Decide which one you are actually testing before you start, because they need different evidence.
The sample is the first real test
Measure the bore at both ends, not just the middle. A sleeve that is round at the centre and oval at the gate end will feel fine by hand and leak on the line.
Check wall thickness at four points around the circumference. Uneven wall is the most common moulding defect in thin sleeves and the hardest to see without a gauge.
Look at the gate area and the parting line for flash. A little flash is cosmetic; flash inside the bore reduces the clearance you designed for.
Compare the colour against your reference under daylight, not under the office light the sample was photographed in.
Then do the only test that matters: seat it on the terminal you actually use, by hand, ten times. Note whether it goes on the same way each time.
Ask how the part is made, not just what it is made of
Moulded and extruded sleeves are different products with different tolerance behaviour. A moulded part holds a designed shape and a length; an extruded one is cut to length and has a different wall consistency. Both are legitimate — they are not interchangeable.
The follow-up question is who owns the tool. If you paid for it and the answer is not a written statement naming you as owner, you are renting access to your own product. This single point decides whether you can change supplier in two years.
Read a test report for what it covers
A compound report describes the material batch. It does not describe the part that came out of the tool, and it does not cover wall thickness, seating force or dimensional tolerance.
Check four things before accepting any report: what sample is identified in it, which parameters were actually measured, which laboratory issued it and when, and whether the standard cited applies to a moulded part or only to a raw material.
A report that answers a different question than yours is worse than no report, because it creates the impression that the question was answered.
Design a first order that can fail
An order with no acceptance criteria cannot fail, which means it also cannot tell you anything. Write down the two or three dimensions that matter for your assembly, and the window you will accept.
Measure ten pieces from the same batch rather than one, because one piece shows the best case and ten show the spread. If the spread is already at the edge of your window on the first batch, the process has no margin left for the second.
What to do when the first batch is wrong
Handle it as a measurement problem before it becomes a commercial one. Photograph the parts against a ruler, state the measured value and the value you expected, and send both in the same message. A complaint without a number invites a reply without a commitment.
Then separate the two possible causes: the tool is wrong, or the process drifted. If every piece is off by the same amount, the tool or the drawing is wrong and no amount of re-running will fix it. If the pieces scatter, the process is unstable and the moulding parameters are the place to look.
How a supplier responds to this message is the most useful verification data you will get, more than any audit, because it shows the behaviour you will live with for the life of the part.
A remote audit covers more than buyers expect
- Photographs of the tool, the cavity and the machine, not just the finished part.
- A short video of a production run, so you can see cycle time and how parts are handled.
- Dimensional data on ten pieces, with the gauge identified.
- The material purchase record, with the grade name — not just PVC or silicone.
- A named contact who answers a technical question directly rather than forwarding it.
Red flags
The one that matters most in practice is the last one. A substitution is rarely malicious; it is usually a material grade that was easier to buy that week. But a part described as equivalent without a stated difference is a supplier telling you that your specification is negotiable.
- The same sample photographs used for several different sizes.
- A certificate with no sample identification and no test date.
- Tolerance given as a range wide enough to cover any outcome.
- Reluctance to state who owns the tool.
- A different part arriving than the one approved, described as equivalent.
Frequently asked questions
Is a test report proof that the parts are compliant?
Only for the sample identified in it and the parameters listed. A report on the compound does not prove anything about the moulded part's dimensions or wall consistency.
Can I verify a supplier without visiting?
Largely yes. Tool photographs, a run video, and dimensional data on ten pieces from one batch cover most of what an audit would tell you about capability. What you cannot check remotely is how they behave when something goes wrong.
Who should own the tooling?
You should, if you paid for it. Ask for a written statement to that effect before the first production run, because after the tool is worn in, the conversation gets harder.
How many samples should I inspect?
Enough to see variation. One sample tells you the best case; ten from one batch tell you the spread, which is what your assembly will actually experience.
Do I need third-party inspection on a first order?
For a custom part, a dimensional check by an independent lab is inexpensive relative to the cost of a wrong batch. For a standard size that a stockist holds, a sample check is usually enough.
What if dimensions drift between batches?
Agree the measurement method and the acceptance window in writing first. Drift without a baseline is an argument you cannot win, because neither side can show what changed.