Writing an Inspection Plan for a Mixed Silicone Accessory Shipment
A shipment that mixes rings, toning balls, mats and blocks looks like one delivery and behaves like five. Each item has its own mold, its own nominal weight and its own failure modes, and the moment a single inspection plan is applied across the whole container, two things happen: light items get sampled as if they were heavy ones, and visual marks that are normal on cast silicone start rejecting entire cartons. The plan has to be written before the molds run, not when the goods are already on the floor.
Fix the weight tolerance before the first casting
Cast silicone does not hold weight the way a machined part does. Fill volume in the mold cavity, cure shrinkage, pigment load and the amount of flash trimmed at the parting line all move the finished weight by a few grams, and the spread is wider on thick sections than on thin ones. That is why a tolerance written as a relative figure is a poor instruction: the same relative band means half a gram on a small block and several grams on a heavy ball.
State it instead as an absolute band per model, tied to the nominal weight on the artwork and the packing list. A hand-held ring can reasonably sit within a band of a few grams, a loaded toning ball within a slightly wider one, and a mat or a foam block should not be judged on the same numbers at all. The accessory range already groups items by construction, so the tolerance table can follow that grouping instead of inventing a new one.
Calibrate the scale, then trust the number
Weight calibration is the step that quietly decides whether the whole plan means anything. A production scale has to be checked with certified test masses at the start of each shift and after any battery change, mains interruption or move between lines. The check mass should sit close to the working range, not at the top of the scale's capacity, and the recorded result has to include the deviation, not just a signature.
At the bench, tare the carton and weigh the net content, then record gross, tare and net per carton rather than a single figure. Keep the resolution of the scale finer than a fifth of the tolerance band, or the measurement error will consume the tolerance on its own. The distinction between attributes and variables sampling explains why a measured weight carries more information per unit than a pass or fail judgment, and why weighing is worth the extra minute.
Stratify the mixed order before applying any acceptance number
A sampling scheme assumes the lot is one population. A mixed shipment is not. Split it into strata by model, by colour, by size and by production date, and let each stratum carry its own sample size and its own acceptance number. If a stratum holds only a few cartons, either inspect it fully or accept on zero defects without an allowance — an allowance applied to a stratum of two cartons accepts almost anything.
Attributes plans work for what can only be judged yes or no: a tear, a missing insert, a wrong colour. Variables plans work for weight and dimension, where the reading itself is the result. Using an attributes plan for weight throws away the number and turns a drifting batch into a pass, which is exactly how a shipment arrives heavy at one end of the carton and light at the other.
Define critical, major and minor before the run starts
The most common cause of a rejected batch is not a defect. It is a defect list agreed after the goods were made, when the buyer's team and the factory are looking at the same carton with different definitions. Write the three classes into the plan and attach a limit sample for every class that is judged by eye.
- Critical: rigid core exposed or detached, a tear that reaches through the wall, a chemical odour that survives aeration, any item that cannot be used safely.
- Major: weight outside the band for its model, a dimension outside the drawing, an insert that sits tilted or proud of the surface, a colour outside the approved standard.
- Minor: mold parting line flash, polishing marks, dust or a faint flow line that does not change grip, function or hygiene.
Cast silicone almost always shows something at the surface. Without a limit sample and a written allowance for minor marks, an inspector who is doing the job properly will still reject pallets, and the argument will be about taste rather than about fitness for use.
Measure the core inside the soft part
On an item built as a soft silicone body over a rigid insert, the outside can look perfect while the inside is wrong. Check insert position against the drawing on a sectioned sample from each cavity, check the thinnest wall over the insert, and confirm that the insert does not rotate or pull out under a defined load. The process controls behind this decide whether the check is a formality or a real gate, and our notes on accessory testing set out the readings worth recording per batch.
Keep the plan to one page and one signature
A plan that runs to six pages is not used on the floor. One page holds the strata, the sample sizes, the three defect classes, the weight bands per model, the scale calibration record and the name of the person who signs the release. The factory signs before the container is loaded, the buyer countersigns on arrival, and any deviation is written on the same sheet rather than in a separate mail thread. If a shipment is contentious, the page is the only document that settles it — send us the model list and the drawing tolerances and we will return a draft plan matched to the molds.
Frequently asked questions
What weight tolerance should I specify?
An absolute band per model, not a relative one. A hand-held ring can sit within a few grams and a loaded ball within a slightly wider band, and each band should be measured on a scale whose resolution is at least five times finer than the band itself.
How is AQL applied to silicone products?
Split the order into strata by model, colour and production date, then apply an acceptance number to each stratum separately. Use attributes sampling for visual defects and variables sampling for weight and dimension, because a measured reading catches drift that a pass or fail count hides.
Do you calibrate every piece or sample the batch?
The scale is calibrated on a schedule with certified test masses; the pieces are weighed individually within the sample. Every unit in the sample gets a reading, and the batch is accepted or held on the distribution of those readings, not on the average alone.