Custom Silicone Parts from China: The Four Things That Decide Whether the Batch Fits

I have sat through maybe forty silicone part projects over the years, and the pattern is depressingly consistent. The first samples are beautiful. The third production batch has flash on the sealing face, the durometer drifted two points, and the part smells like vinegar when you open the bag. All three are avoidable, and none of them are cured by asking the factory to "try harder".

Start from the application, not the drawing

Silicone gets specified casually because it is forgiving. It stays flexible from -50 to about 200 degrees Celsius, it plays nicely with food contact and skin, and it does not embrittle with age the way some cheaper elastomers do. That tolerance is exactly why people stop thinking.

Before quoting, be clear on whether the part is a static seal, a dynamic one, a grip, a gasket that will see hot oil, or something a baby will chew on. A static cover gasket at Shore 50A and a valve seat that flexes ten thousand times are not the same project, and a factory that quotes them identically either did not read your drawing or is planning to send you whatever compound was already in the mixer.

Durometer is a range, not a number

You will often see "50 Shore A" on a drawing as if it were a machining tolerance. In practice, molded silicone hardness is controlled by the filler loading in the compound and by cure, and a realistic production window is plus or minus three to five points. If your design only works at exactly 50A, it will not work.

Two things help. First, specify a range and test against it with a durometer on a flat, thick-enough section, not on a curved surface where the reading is meaningless. Second, ask for the actual batch reading recorded per lot. A molder who cannot produce a durometer log for the lot has not measured anything.

If you are designing something new, run a quick hardness sensitivity check at 45A, 50A and 60A on your prototype. Plenty of teams discover their seal only works at the top of the window, which is a very cheap discovery before tooling and a very expensive one after.

Platinum cure or peroxide cure, and why the smell matters

Most custom parts get platinum-catalyzed (addition cure) silicone these days, because it is cleaner, has less shrinkage variability and does not leave the faint acetic or peroxide odor. Peroxide cure is cheaper and still fine for some industrial parts, but it can leave byproducts that are a genuine problem for medical, food-adjacent or enclosed electronic applications.

There is a well-known wrinkle with platinum cure: sulfur, tin, amine-cured epoxy and some clays will poison the catalyst, causing tacky surfaces or incomplete cure. If your part has to cure in contact with an insert or a fixture that was molded with a different material, say so, and confirm the factory handles it - good shops keep dedicated tooling and dedicated space for platinum work.

Tooling: venting and parting line, not just cavity count

Silicone flows and cures differently from thermoplastics. Air has to leave the cavity, so venting design is what separates a clean batch from a deflashing nightmare. Ask where the vents are and how they are maintained. Vents clog with cured silicone over a production run, and a shop that never cleans them will progressively produce worse parts as the order goes on.

The parting line position is the other thing to approve on the tool drawing, before the tool is cut. On a sealing face, a parting line that lands on the sealing surface is a leak path. Move it to a non-functional surface, and accept that the tool may cost a little more.

Tool ownership deserves a sentence in the contract. Pay for it, put your name on the steel, and ask for a photo of the tool with your identifier stamped on it at first sample approval. Otherwise, when you move to another supplier in three years, the conversation gets awkward.

Tolerance reality check

Silicone shrinks, and the shrinkage varies with wall thickness and cure. Expect general tolerances in the region of plus or minus 0.2mm for small dimensions, tightening to maybe 0.1mm only on critical features with dedicated tooling and 100% inspection. Anything quoted as "zero tolerance" is a sales answer, not a manufacturing answer.

As a practical matter, design seals so the compression percentage carries the variation rather than the geometry. A groove that gives you 20 to 30 percent compression has room for real-world tolerance. One that relies on 5 percent does not.

What to put in the purchase order

Compound reference and cure system, stated explicitly, with a hardness range.
Color reference and a signed color chip from the first approved lot.
Parting line location and allowed flash.
Any post-cure requirement, since post-curing changes hardness and removes volatiles.
Certificate of conformity per lot, plus durometer readings.
For food, skin-contact or medical claims, the specific test report and whose name is on it.

That last point deserves emphasis. A factory holding a compound supplier's FDA or LFGB declaration is not the same as the molded finished part being tested. If the claim matters commercially, get the report for your part, your compound, your lot, and check the report actually names the material and the test method.

The first article is not the last word

Approve the first article with a dimensional report, a durometer map across at least five parts, and a visual against your flash standard. Then check the third production lot against the same numbers, without telling the factory in advance which lot you will pull. Silicone is stable once the process is under control, but the process only stays under control if somebody is watching.

None of this is exotic. It is just specification work, done before the tool is cut rather than after the parts arrive with a leak.