A receiving inspector measures a sample from a lot of molded silicone parts. Most pieces are inside the drawing limits. A few are not. The first instinct is to call the whole lot nonconforming, or to loosen the tolerance so the lot passes. Both reactions can hide the real cause.
With molded silicone, an isolated deviation can come from at least three different places: the part came from a particular mold cavity, the lot differs from earlier lots, or the two sides are not measuring the same way. Each one calls for a different conversation with the supplier. This article is a short guide to telling them apart before the argument starts.
Start by asking where the failing parts came from
If a mold has more than one cavity, parts from different cavities can differ slightly even when the mold is built to the same drawing. Whether you can see this depends on whether the parts are identifiable by cavity at all.
Before you open a dispute, check three things:
- Do the parts carry any cavity mark or other identification, or was that never specified?
- Were the failing pieces packed together, or spread across the lot?
- Does the supplier’s lot record, if it exists, link a pack to a cavity or a molding run?
If the failing pieces cluster in one pack or one marking, the question becomes a tooling question. If they are scattered and unmarked, you cannot say that yet, and it is better to say so than to guess.
This is also a reason to decide early whether cavity identification matters for your part. For a part with a critical sealing or fit feature, it is much easier to ask for identification in the original RFQ than to ask for it after a lot has shipped. Whether it is practical for a given mold is something the supplier has to confirm for the specific project.
Compare against earlier lots, not only against the drawing
A drawing tells you the limit. It does not tell you whether this lot moved. If you kept data or a retained piece from the first batch, set the failing parts next to it. Two outcomes are common:
- The deviation is new in this lot. That points toward a change in material, process or tooling condition, and the supplier should be able to say whether anything changed.
- The deviation was present in earlier lots too, and nobody noticed because the first inspection was less detailed. In that case the issue sits in the drawing, the tolerance or the original agreement, not only in this lot.
Both are legitimate findings. They lead to very different next steps, which is why an undocumented “it looked fine before” is not enough to work from.
For the retained piece, remember what it is. A sample kept on a shelf can change with time and handling, so treat it as a comparison aid, not as the design dimension. The reference sample note on this site goes into how to identify one by revision.
Check that both sides measured the same feature the same way
Many “failures” turn out to be measurement differences. Silicone is soft, so the reading can change with contact force, with where on the part the gauge touches, with how long the part rested after molding, and with the temperature and handling of the sample. If the buyer measured with a contact tool and the supplier with a non-contact method, a small gap between the two numbers is not surprising.
Settle these before comparing numbers:
- Which feature is being measured, and from which datum on the drawing.
- Which instrument or method each side used.
- How the part was held and how long it was left to settle.
- How many readings were taken, and how they were rounded.
The measurement method article linked here covers what to put in an RFQ so that this agreement exists before the first lot. If it does not exist yet, writing it down now is part of resolving the deviation.
Decide whether the feature is actually functional
Not every dimension on a molded part controls fit or sealing. If the out-of-limit feature is a non-functional surface, the right conversation may be about relaxing or reclassifying that dimension on the drawing, not about rejecting the lot. If it is a sealing or assembly feature, the bar is higher, and it is worth stating what happens in the assembly when the dimension is off. Marking the functional surfaces on the drawing is the cleanest way to stop this from recurring.
What to send the supplier
A short, factual message is more useful than a rejection notice. Include:
- Drawing revision and the feature that is out of limit, with its datum.
- How many pieces were checked and how many failed.
- Any identification on the failing pieces, and how they were packed.
- Your measurement method and conditions.
- Photos of the failing pieces next to a piece that passes.
- Whether an earlier lot or retained sample exists for comparison.
With that, the supplier can check cavity, lot and method, and tell you what can and cannot be determined from the information available. Corrective action, replacement or a drawing change are decisions to make afterwards, once the cause is known.
If you are still at the RFQ stage
The cheapest time to prevent this is before the first order. Ask how lots and cavities are identified, agree the measurement method, mark the functional features, and decide what a first-batch check will look like. Our molded parts, first batch and lot separation notes cover those inputs.
If you have a molded part drawing, or a lot that you are trying to understand, you can send it for engineering review. What is feasible, and how it would be checked, is confirmed for the specific project.