Quality
22 October 2026
The second order arrives, the navy looks purple, and the argument starts. In most of these cases nothing went wrong in production. The buyer approved a colour from a phone photo of a chip, and a photo is not a standard — it is a rendering of a standard through a camera, a screen and a colour profile that nobody recorded.
Colour has to be approved as a physical object under stated light, against a stated number. Here is how that works for moulded cases.
ΔE is a single number describing how far apart two colours are in a perceptual colour space. It is not a percentage and it does not have an absolute meaning; what matters is the scale and the material. For moulded thermoplastics the working bands look like this:
| ΔE (CIEDE2000) | What it looks like | Typical use |
|---|---|---|
| < 0.5 | Indistinguishable side by side | Two-shot parts that must match |
| 0.5 – 1.0 | Visible only in a trained comparison | Premium retail, tight tolerance |
| 1.0 – 2.0 | Noticeable on inspection, acceptable in use | Normal commercial tolerance |
| 2.0 – 3.5 | Clearly a different batch | Accept only if the buyer signed it off |
| > 3.5 | A different colour | Reject |
Write the number into the purchase order. A PO that says "as approved sample" without a ΔE figure leaves the decision to whoever is holding the carton.
The same case looks navy under daylight, blue-black under an office fluorescent tube and almost purple under warm LED. This is metamerism plus a genuine pigment effect, and it is why colour arguments often cannot be settled in a room with one light source.
Standard practice is to judge under three sources: D65 (north daylight), TL84 (European retail fluorescent) and A (incandescent). A colour that holds across all three is a stable colour. A colour that only holds under one is a colour that will generate a complaint from a customer standing in a different shop.
Most real deviations come from four causes, none of them dramatic: a different pigment lot from the masterbatch supplier, regrind content creeping above the agreed percentage, a barrel that was not purged between colours, and a change of texture grain on the mould. The first two are the frequent ones, and both are invisible in a photograph.
If a repeat order must match a previous one, say so in the PO and ask for the retained chip to be pulled and re-measured before the run starts. That one sentence prevents most of the conversations we end up in.
Material choice affects how stable a colour is over a product's life; the material comparison explains why silicone drifts and TPU does not.
1.5 to 2.0 under CIEDE2000 is a normal commercial tolerance for moulded thermoplastics. Tighter than 1.0 is achievable but raises the rejection rate and the price.
No. A photo is a rendering of a colour through a camera and a screen. Approve a physical chip in the final material and finish, signed by both sides, under a stated light source.
Metamerism plus the light source. Judge under D65, TL84 and incandescent; a colour that only holds under one source will generate complaints in retail.