The plastic shrinks away from the steel as it cools, and shrink is not a constant — it moves with the resin, the wall thickness, the gate, the pack pressure and the mold temperature. The cavity dimension is cut once and stays put. The part dimension is the cavity minus a number that varies shot to shot and grade to grade. Quoting the steel figure as though it were the part figure is the most common way a tolerance conversation goes wrong, and it goes wrong in the supplier's favour, which is why it is worth being blunt about.
If both ends of a dimension are formed by the same tool detail, the only thing between the drawing and the part is shrink. Nothing has to line up, because nothing moves relative to anything else. When a dimension genuinely has to be held, the first question is whether it can be moved onto a single detail — that is a design change, not a process one, and it is free while the part is still in CAD.
A dimension measured from one half of the mold to the other carries the alignment of the two halves, and that alignment moves with clamp force, thermal growth and wear. It is a real and unavoidable addition to the variation, and it is why an otherwise identical dimension is looser across the split than within one half.
A moving detail has to locate itself on every cycle, and it does so against a positive stop that wears. Features formed by a side action or a lifter inherit that repeatability on top of everything else. This is the hidden cost of an undercut: not only the tooling, but a permanently wider tolerance on whatever the moving steel makes.
Every resin shrinks across a range rather than at a number, and filled grades shrink several times more across the flow than along it. On a tight dimension that range alone can consume most of the drawing tolerance before the machine, the tool or the gauge have contributed anything. That is a calculation worth doing before a tolerance is committed, not after a first article fails.
Check it — the shrinkage calculator shows how much of your band the resin takes
Gauge variation is part of the budget, and at small dimensions it is a large part — a contact probe deflects the thin wall it is measuring, and optical methods bring their own edge-detection uncertainty. A GR&R study tells you how much of the band the measurement itself is eating. Committing to a number without that is committing to an argument later about whose gauge is right.
There is no single number for "standard" injection molding tolerance, which is why searching for one returns a dozen incompatible charts. What exists is a set of systems that grade tolerance by dimension size and by how demanding the feature is.
Current
The German standard in force for plastic moldings. It assigns tolerance groups and scales the allowance with the nominal dimension, so a tolerance that is routine at 5 mm is not the same allowance at 100 mm. It also separates dimensions the tool controls directly from those affected by the parting line or moving steel — the same distinction this page makes.
Withdrawn — replaced by DIN 16742
Still cited constantly on old drawings and still searched for more than its replacement. It was withdrawn, and 16742 supersedes it. The two are not a line-for-line mapping: the group structure changed, so a drawing calling out a 16901 group needs converting rather than renaming. If you have inherited a print citing 16901, that conversion is worth doing explicitly instead of assuming the nearest-looking group.
Current, international
The ISO equivalent for plastics moldings, structured on the same idea of tolerance grades tied to nominal size. Where a program needs an international rather than a German reference, this is usually the one.
Guideline, not a standard
The commercial / fine / precision bands common in North America. Useful shorthand for a first conversation, and not a specification — they are trade-association guidance rather than a controlled document, and a drawing should not cite them as though they were.
Deliberately no tolerance chart here. The groups belong to the standards themselves, and a table copied from memory is how wrong numbers end up on a drawing under a standard’s name. If you have a dimension that has to hold, send the part and we will tell you which side of it the steel can reach. If the question is the other way round — whether the tolerance you need is one anybody can hold — the Comparator puts it in context.