Most tolerance disputes on plastic parts do not start with a bad mould. They start with two people reading different standards — or reading the same standard differently. The current reference is ISO 20457:2026, Plastics moulded parts — Tolerances and acceptance conditions; the 2018 edition has been withdrawn. This page explains the framework: where it came from, how a tolerance grade is derived in principle, what W and NW mean, and what has to be agreed before anyone measures anything.

This article describes the framework and does not reproduce the standard’s tolerance tables, which are copyright-protected.

A drawing may legitimately cite any of these, but they are not interchangeable: values for the same nominal size and grade differ between editions. Choose one framework per drawing and state its edition. For new work, cite ISO 20457:2026.

• TG tolerance grades, tightest to loosest. The grade is not looked up by resin name.
• W (tool-related) vs NW (non-tool-related) dimensions, with separate values — NW is generally wider.
• General tolerances expressed as surface profile tolerance within a specified datum system for all integral features, with additional ISO-GPS dimensional and geometrical tolerancing where function requires it.
• Acceptance conditions written into the standard — conditioning, ambient conditions and measurement method are part of acceptance.
• Scope: injection moulding, injection compression, transfer, compression and rotational moulding of non-porous parts from thermoplastics, thermoplastic elastomers and thermosets; other processes by agreement. Surface imperfections such as sink marks, undesired flow structures, roughness and joint lines are not addressed. It is not intended to supplant tolerance requirements in product standards.

The grade comes from an evaluation of the production situation rather than a material lookup. The standard scores a set of factors — such as the production process, the stiffness or hardness of the moulding material, the calculated moulding shrinkage, the shrinkage difference caused by geometry and process, and the production and QA effort invested — and the combined result determines the grade.

Two practical consequences:

1. The same resin can land in different grades. Unfilled PA6 and PA6-GF30 differ in stiffness, shrinkage and anisotropy, so they differ in achievable grade. A designer who only states “PA6” has not given the supplier enough to quote against.
2. The grade is partly a commercial statement. The production/QA factor rewards suppliers who invest in process control. If a supplier will not discuss how they justify the grade they propose, ask why.

• W (tool-related): the dimension is formed by a single mould component — e.g. a bore entirely within one core.
• NW (non-tool-related): the dimension is formed by the interaction of moving mould parts — across a parting line, or affected by slides and inserts. Wall thickness measured across a parting line is the classic case.

NW values are wider than W values for the same grade and size. Classifying a feature incorrectly is one of the most common ways a “tight tolerance” becomes unachievable — or, conversely, how a supplier over-promises. Classify each critical feature on the drawing, not in an email afterwards.

Agree in writing before the first shot:

• Conditioning and environment — measurement in a standard atmosphere (commonly 23 °C / 50 % RH) is the long-standing convention. Specify your own tolerance band on temperature and humidity, a minimum conditioning time after moulding, and a maximum window, so both parties measure the same material state.
• Method — CMM vs dedicated fixture, datum scheme, clamping force, probe strategy.
• Non-rigid parts — datum target points should be defined, and gravity and fixturing effects considered.
• Verdict rule — agree what happens when a general tolerance is missed but function is unaffected, and which dimensions are verified.

If supplier and customer measure at different times, temperatures or with different fixtures, both can be correct and still disagree. That is a process failure, not a mould failure — and the acceptance conditions decide who is right.

If your parts are made in China, note that GB/T 47067-2026 is a modified adoption of ISO 20457, and ISO has since published a newer edition. A supplier working to the Chinese standard and a European customer citing ISO 20457:2026 may therefore be working to different editions of the same framework.

In most projects this is harmless — the TG structure, the W/NW logic and the acceptance principles belong to the same family. It becomes a problem when a drawing cites one edition while the supplier’s inspection instructions cite another, and a dispute turns on a value that differs between them.

How to close the gap: state one framework and one edition on the drawing, and have the supplier confirm in writing which standard their inspection procedure actually follows. Ask them to confirm equivalence for your critical dimensions rather than assuming it.

A note like this removes most of the ambiguity discussed above. Adjust the bracketed values to your own agreed conditions:

General tolerances per ISO 20457:2026, grade [TGx], W/NW classification as marked. Critical dimensions individually toleranced as shown. Parts measured at [23 ± 2 °C / 50 ± 10 % RH] after a minimum of [X h] conditioning and within [Y h] of moulding, using [CMM / dedicated fixture], datums as defined. Any deviation from these conditions to be agreed in writing before measurement.

The tolerance value tables are the core content of the standard and are copyright-protected. Obtain them from ISO for the current edition, from DIN for the German adoption (check the DIN catalogue for the current version), or from China’s national standards platform for GB/T 47067-2026.

Do not build quotations or acceptance criteria on tables copied from third-party websites. Transcription errors in circulated versions are common, and a wrong number in a contract is far more expensive than the cost of the standard.