Boundary Sample Approval — How Do You Set Upper and Lower Limit Samples to Define the Acceptable Quality Range for Your Injection Molded Parts?
Last updated: 2026-09-03 · Reading time: about 6 min
A boundary sample — also called a limit or golden sample — is a signed physical part that fixes the acceptable range for a characteristic a drawing cannot fully capture: color, gloss, texture, sink or weld-line appearance. Mold 50–100 parts at the production process, sort them, and sign an upper, a lower and a golden sample with your customer before mass production.
See also related article.
Key facts
<>
Shrinkage — The dimensional reduction from mold cavity to cooled part. Amorphous resins: 0.4–0.7% (uniform). Semi-crystalline: 1.0–2.5% (directional). Mold makers apply shrinkage factor when machining cavity. Actual shrinkage varies with wall thickness, gate location and packing pressure. See also related guide.
Which characteristics need a boundary sample — and when is a drawing not enough?
A drawing fixes a number. Some traits cannot be reduced to one number the eye will accept — those need a physical limit.
| Characteristic | Why a drawing is not enough | Sample type |
|---|---|---|
| Color | ΔE ≤ 1.0 passes on paper, but two parts at ΔE 0.8 can still look different if the shift runs in the hue direction | Upper + lower (color chip) |
| Gloss | Read in GU at 60°, but gloss varies across the surface — lower at the gate than the far end — and the drawing does not say where to measure | Upper + lower (flat plaque) |
| Texture / etch depth | Depth in μm (e.g. 30–50 μm), but appearance shifts with resin color, draft angle and lighting | Upper + lower (textured plaque) |
| Sink mark visibility | Depth in μm, but visibility depends on lighting and viewing angle — a 20 μm sink shows on glossy black, hides on textured gray | Upper only (worst acceptable sink) |
| Weld line | Strength in MPa, but the visible line on the surface is subjective and lighting-dependent | Upper only (worst acceptable appearance) |
Use a boundary sample when the trait is visual, when the drawing tolerance is looser than the assembly really needs (the drawing says ±0.1 mm, the fit needs ±0.05 mm), or when a visible defect was never specified. Skip it where a dimension is fully toleranced on the drawing — the CMM measures it — or where a functional test returns a number rather than a visual call.
How do you create and sign off boundary samples — step by step?
Four steps, from a stable process to a stored, signed reference set.
- Mold 50–100 parts at the production process. Run them at the T2 or T3 trial — the same injection speed, melt and mold temperature and packing pressure mass production will use. Samples molded at a manual T1 process do not represent production parts, so the boundary would be invalid.
- Sort the parts by the characteristic. Measure or evaluate all 50–100 and rank them — lightest to darkest for color, shallowest to deepest for sink. The spread from best to worst is the natural process capability, and both limits must sit inside it.
- Select the upper, lower and golden samples with the customer. The customer reviews the sorted parts and picks the worst acceptable part on each side plus the ideal target. Both sides agree together — neither the mold maker nor the customer sets the limit alone.
- Sign, date and store them. Label each sample with the part number, mold number, characteristic, limit type, date and both signatures. Store at 20–25 °C and 40–60% RH so it cannot fade, warp or degrade; keep one set each for the customer, the mold maker and the production floor.
A boundary sample does not replace the CMM — the two are used together. The CMM verifies dimensions objectively, a number against a tolerance; the boundary sample verifies what the eye and hand judge — color, gloss, texture, sink. A production part must pass both: correct dimensions but off-limit color is a reject, and correct color with an out-of-tolerance dimension is a reject too.
“A drawing gives you a number; a boundary sample gives you what the customer will actually accept on the floor. We mold the limit set at the production process, sort it, and sign the upper, lower and golden samples together with the customer before mass production — so nobody argues over a shade of gray at 100,000 parts.”
— Project Engineer, aCatMould
Frequently asked questions
What happens when a boundary sample degrades — and how often do you replace it?
It drifts: color fades under light, the surface scratches from handling, and the resin absorbs moisture.
Replace a sample when color moves past ΔE 0.5 against the master (read on a spectrophotometer), when the surface is scratched enough to distort the visual call, or when the sample is lost.
To replace it, mold 10–20 parts from current production, pick the one that best matches the master, and have the customer sign the new sample. Mark the old one SUPERSEDED and keep it for reference only. Check the working set against a controlled master every 6 months.
Can you use a boundary sample for a functional dimension — or only for visual traits?
Not for a bare dimension. A functional size must be measured by a CMM, caliper or gauge, where the result is a traceable number. Judging a dimension by eye against a sample is subjective — two inspectors disagree — and untraceable, since the sample cannot be calibrated.
The exception is a go/no-go gauge: a physical device the part either fits or does not. It is a functional boundary in hardened form, calibrated against a master gauge with a certificate, and it checks a dimension in 1–2 s versus 30–60 s on a CMM — so it suits high-volume inspection.
Do you need a boundary sample for every characteristic on the part?
No. Only for traits a drawing cannot pin to an acceptable number — the visual and tactile ones. Fully toleranced dimensions go to the CMM, and functional tests that return a number (pull force, seal pressure) go to the gauge or test rig.
A typical set covers two to four cosmetic characteristics — color, gloss, texture, and one defect such as sink or weld line. Each extra signed sample adds storage and re-check overhead, so limit the set to what the eye actually judges at incoming inspection.
Key takeaways
- A boundary sample is the eye’s tolerance — a signed upper, lower and golden part for traits a drawing cannot reduce to one acceptable number: color, gloss, texture, sink and weld-line appearance.
- Create it at the production process (T2/T3): mold 50–100 parts, sort them, select the three limits jointly with the customer, then sign, date and store at 20–25 °C / 40–60% RH.
- It complements the CMM rather than replacing it — a part must pass both — and it needs upkeep: re-check every 6 months, replace when color drifts past ΔE 0.5 or the surface is scratched.
A boundary sample turns a subjective argument into a signed reference both sides agreed to before production. It is how a mold maker and a customer look at the same part and the same call — not by opinion, but by comparison to a limit chosen, dated and stored on purpose. If your part has a cosmetic surface, settle the upper, lower and golden samples during the DFM and trial stage, before the first full production run.
Sources
BYK Instruments — color (ΔE) and gloss-unit measurement practice
Zeiss Industrial Metrology — CMM dimensional inspection
Mitutoyo — measurement instruments and go/no-go gauging
Mold-Tech — mold texture and etch-depth standards
Plastics Technology — cosmetic defect and appearance acceptance practice
Sources: Plastics Technology · Plastics Technology Online · Society of Plastics Engineers (SPE) · ThomasNet