A supplier tells you their process holds Cpk 1.67. That single number can mean “we run a stable, well-controlled process” — or “we measured 30 parts off a freshly tuned machine on a good day.” Here is how to tell which one you are looking at — written from the shop-floor side at TONGDA LINK, where we run these studies on medical and precision tooling every week.

→ For realistic tolerance ranges on plastic parts, see our guide on

• Cp = (USL − LSL) / 6σ — spread only. It ignores whether the process is centred.
• Cpk = min[(USL − μ)/3σ, (μ − LSL)/3σ] — spread and centring. This is the number that matters.
• Pp / Ppk — same maths, but using overall (long-term) variation rather than within-subgroup variation. Ppk is normally used for initial process studies, before the process is proven stable.

A part can have excellent Cp and poor Cpk: tight spread, wrong mean. On plastic parts this is common, because shrinkage pulls the mean away from nominal — and shrinkage is a range, not a constant.

These are industry conventions (AIAG / IATF practice), not universal law — the customer’s specific requirement governs.

1. Sample too small or too convenient. A capability index from 10–20 parts is not a capability index. Ask for the sample size, subgroup structure and time span.
2. Only the “good” cavity reported. On multi-cavity tools, cavities differ. Ask for Cpk per cavity, not an average across cavities.
3. Measured under favourable conditions. Parts measured hot, straight off the machine, or without agreed conditioning will look better than parts measured at the incoming inspection.
4. No MSA behind it. If gauge repeatability is poor, the Cpk is measuring the measurement system, not the process. Ask whether GRR was performed.
5. No re-validation after change. Material lot, resin supplier, machine, tool repair or a cavity polish all change the distribution. Ask how often capability is recalculated.

For each critical dimension: Cpk per cavity, based on at least 30 consecutive shots across 3 production runs, measured at 25 ± 2 °C / 50 ± 10 % RH after 24 h conditioning, using CMM, with X̄–R control charts, measurement method and gauge R&R results attached.

Metal parts are dimensionally stable once machined. Plastic parts drift: post-moulding shrinkage, moisture uptake in PA/PC/PBT, continued crystallisation, and stress relaxation all move dimensions after ejection. A process that shows Cpk 1.67 in week one can degrade as the tool wears or as the resin lot changes. So ask for capability over a sustained run, and for the plan to re-verify it.

See also: why plastic part dimensions change after 48 hours.

For medical-device customers, TONGDA LINK does not report a single headline Cpk. We report per-cavity, per-feature data with the conditions attached, because a number without conditions is not auditable.

Typical results. On critical features of single-cavity and multi-cavity tools running for medical assemblies, our per-cavity Cpk values most commonly fall in the 1.33 ≤ Cpk < 1.67 band, with some features exceeding 1.67. We do not claim a blanket figure across all parts: the achievable index depends on the nominal dimension, the tolerance band, the resin, and the measurement method. Where a customer requires Cpk ≥ 1.67 on a given feature, we agree that target before cutting steel and design the gating, cooling and tolerance allocation accordingly — or tell you honestly that the tolerance needs to be relaxed.

The following is our standard deliverable set for regulated programmes; items marked required ship with every medical-tool FAI package at TONGDA LINK:

FAI first (required) — actuals per cavity and per feature, not pass/fail only.

Cpk per cavity (required) — not pooled across cavities, with sample size, subgrouping and run dates shown.

X̄–R control charts (required) from the validation run(s), so you can see stability, not just the index.

Drift study at 24 h / 48 h / 72 h (required) on the same samples, with deltas recorded — because a Cpk taken at 2 hours does not describe the part you receive at day three.

Measurement conditions written in (required) — ambient temperature/humidity, conditioning time, fixture or CMM, datum scheme and probe strategy.

MSA / gauge R&R (required) for each critical characteristic, so the index reflects the process, not the gauge.

Report ownership — each report carries tool ID, cavity map, run dates, operator and measuring-equipment ID, so an auditor can trace it back to the shot.

Cpk describes a dimension. It does not describe whether the tool, the machine and the process are locked down. For regulated programmes we support full process validation structured as IQ, OQ and PQ — the same three-stage logic your quality team uses for production equipment, applied to the mould and its moulding process.

IQ — Installation Qualification. Documented evidence that the tool and its installation match the design intent. Typically covers: tool identification and cavity count; steel and heat-treatment certificates; verification of slides, lifters and ejector stroke; thermocouple and hot-runner controller calibration records; machine tonnage, platen and nozzle compatibility; and confirmation that all sensors, interlocks and safety devices function as specified.

OQ — Operational Qualification. Establishing the working window and proving the process is repeatable within it. This includes a process-window study over melt temperature, mould temperature, injection speed/pressure, hold pressure and time, and cooling time; determination of upper and lower limits; generation of limit/boundary samples; and demonstration that the process returns to centre after a normal stop/start.

PQ — Performance Qualification. A sustained production run at nominal settings producing conforming product. Executed as e.g. three consecutive runs of 1000 shots each, with per-cavity dimensional results, Cpk/Ppk on agreed characteristics, visual and functional checks, packaging and labelling verification, and full lot traceability.

Re-validation triggers (change control). Capability and qualification are not one-time events. We define in advance which changes require re-qualification: resin grade or supplier change, material lot change where shrinkage differs, tool repair or cavity polishing, insert or core replacement, machine transfer, and any change to the measurement method.

Documentation. At TONGDA LINK, protocols are approved by the customer’s quality engineer before execution; reports include raw data, deviations and conclusions.