Answers › Part tolerances

How tight can tolerances be on an injection-molded part?

There's no single number — achievable tolerance depends on the part's size, material, and geometry. Because a molded part shrinks as it cools, tolerance is a conversation about what the part must do, not an arbitrary tightness. The practical rule: hold a tight tolerance only where the part needs it, and let the rest float.

Why molded tolerance isn't a fixed number

Every molded part moves a little as the plastic cools and shrinks, and different resins shrink by different amounts — so the "holdable" tolerance changes with the material and the feature. A dimension that mates with another part or seals against something must be controlled closely; a cosmetic edge usually doesn't. Treating every dimension as critical just drives cost up for no benefit.

How to set an injection-molding toleranceA decision tree: decide how tight a dimension must be — commercial-standard for cosmetic features, fine-tight for mates and seals, critical with GD&T for function — then weigh what drives the achievable tolerance: resin shrink, part size, parting-line layout, and tool and process precision.Cosmetic, non-functionalMates, seals or fitsFunction depends on itMaterialSizeTool layoutTool + process

How tight does this dimension really need to be?

Commercial / standard tolerance: cheapest, the default

Fine / tight tolerance: needs better tool, control, cost

Critical: GD&T plus inspection

What drives the achievable tolerance?

Low-shrink resin holds a tighter band

Bigger nominal, wider tolerance

Same-side beats across the parting line

Precise tool and steady process hold more

Setting a molded tolerance in two steps: first decide how tight each dimension really needs to be — a cosmetic feature can sit at a commercial/standard tolerance, a mating or sealing feature needs a fine/tight one, and a function-critical feature gets GD&T and inspection — then weigh what actually makes that tolerance holdable: the resin’s shrink, the nominal size, whether the feature lives on one side of the tool or crosses the parting line, and the tool and process precision.
Relative, qualitative comparison of the three tolerance tiers — directional guidance, not a quote. Published grade tables (the SPI/PIA charts and ISO 20457, which replaced DIN 16742) give the actual bands by feature size and resin.
Tolerance classTypical useCost impact
Commercial / standardCosmetic edges, non-functional walls — the default on most dimensions.Lowest: the band the resin and tool give you for free.
Fine / tightMating faces, bores, sealing surfaces and snap engagements that must fit.Higher: needs more precise tooling, tighter process control and more inspection.
Critical (GD&T)The few dimensions function depends on, referenced to a datum and inspected.Highest: GD&T scheme, a capability study, and sometimes a check fixture or secondary operation.

Use a standard grade, not a guessed number

Rather than inventing a tolerance, molded dimensions are usually specified against published grade tables — the SPI / Plastics Industry guidelines or the ISO 20457 standard (which replaced DIN 16742 and, before it, DIN 16901) — which give achievable ranges by feature size and resin. A few things worth knowing about how they work:

Fine bands and tight tolerance groups are achievable, but whether you can hold one depends on the resin, the feature size, and the molder's process capability — so confirm it rather than assume it.

Critical vs. float: the highest-leverage decision

Tighter tolerances cost more because they need more precise tooling, tighter process control, and more inspection; looser ones let the tool run faster and cheaper. Putting the tight tolerances only where they matter is where the money is.

A worked example (illustrative): on one part you might mark a bore that mates with a pin as critical and reference it to a datum; a snap-fit engagement as critical too; and a cosmetic outer wall as "float" at the general grade. Three dimensions, three different tolerance decisions — which is the whole point: control the two that matter, free the one that doesn't.

One timing point: the tool is cut with the tolerance and shrink compensation built in, so it's largely fixed at build. That's why a first-article (T1) sampling round matters — a molder can remove steel on dimensions deliberately left “steel-safe” to bring them into spec after the first shots, which is cheap compared with re-cutting a finished tool.

What makes a tight dimension easier (or harder) to hold

Why a tighter tolerance costs real money

Every tolerance you tighten past what the part needs adds cost somewhere, and it helps to know where. A tighter number can mean a slower, more controlled molding cycle, more scrap when shots drift outside the band, extra inspection to prove conformance, and sometimes a secondary machining or sorting step the loose version would not need. Tolerances also do not stack for free: on an assembly, the gap or interference you care about is the sum of several part tolerances, so chasing one dimension to a tight number while its mating parts stay loose spends money without fixing the fit. The discipline that keeps cost down is to tighten only the few dimensions that control function — a bearing bore, a sealing face, a snap engagement — and leave the rest at the standard molding tolerance the resin and tool give you for free.

What actually moves a molded dimension

Molded tolerances are not arbitrary; they come from physical sources you can design around once you know them. Shrinkage is the big one — the part is smaller than the steel and shrinks as it cools, and how much it shrinks varies with the resin, the wall thickness, the pressure and the cooling, so a semi-crystalline resin like nylon or acetal holds a looser natural tolerance than an amorphous one like ABS or polycarbonate. Dimensions that cross the parting line or a moving side-action carry the tool's own tolerances plus any clamp or wear movement, so they are inherently looser than a dimension held entirely within one side of the tool. Warpage, moisture uptake and temperature at the time of measurement all shift the number too. Designing the critical dimensions to sit within one side of the tool, away from the parting line and knit lines, is how you earn a tighter tolerance without paying for it.

Standard tolerance tables and when to use them

Rather than invent a number, start from the published injection-molding tolerance guidance (for example ISO 20457, which replaced DIN 16742, or the long-used SPI/PIA tables). SPI gives commercial and fine bands; ISO 20457 uses numbered tolerance groups by resin, applied by dimension size. Put the commercial band on every dimension by default and reserve the fine band for the handful that control fit or function, each one justified. This gives the toolmaker and the molder a shared, hedged reference instead of a blanket ±0.05 mm that may be impossible in one resin and trivially loose in another. On the drawing, call out the critical dimensions explicitly with their tolerance and leave the rest to a general note tied to the standard — a GD&T datum scheme for the features that mate, a general tolerance block for everything else.

Prove the tolerance with a capability study, not a single part

A first-article part measuring in-tolerance does not prove the process will hold it; one part can land anywhere in the spread. The honest way to confirm a tolerance is a short capability study — measure a run of parts, look at the actual spread against the tolerance band, and compute a capability index (Cpk) so you know how much margin the process really has. A tolerance the process can only just hit will produce occasional rejects in production; one with comfortable capability runs clean. This is where a tight tolerance either earns its cost or gets relaxed: if the study shows the process cannot hold the fine band without heroics, the cheaper fix is often to open the tolerance and change the mating part, not to fight the molding process. Our Cpk calculator and tolerance stack-up calculator turn these measurements into the numbers that decide it.

Flexible and thin parts: free or restrained?

A quality and test lab with long benches, drawer cabinets, bench instruments and a cabinet-style test chamber
A quality and test lab at a partner factory. Where and how a part is held during measurement has to be agreed on the drawing, or two labs can disagree about the same part.

A long thin wall, a large lid or a soft TPE gasket can bow under its own weight, then sit flat once it is screwed into the assembly. Measured loose on a surface plate it fails; held the way it is used, it passes. If the drawing does not say which, two inspectors can reach opposite verdicts on the same part.

The molding tolerance guide gives the band to put on each restrained dimension.

Sources for this section:

Common questions

How tight can molded tolerances be? There's no single number — it depends on size, material, and geometry; hold tight only where the part needs it.

Are there standard grades? Yes — specify against the SPI/PIA charts (commercial vs fine) or ISO 20457 tolerance groups rather than a guessed number.

Why do tight tolerances cost more? They need more precise tooling, tighter process control, and more inspection.

Does material/geometry matter? Yes — low-shrink resins (glass fill helps, but can warp) and short, well-supported, uniform-wall features hold best.

How do I put them on a drawing? Tight only on critical dims, referenced to a datum; let the rest sit at a general commercial tolerance.

JCH Design & Manufacturing can review your drawing, confirm what's realistically holdable in your material and geometry, flag anything hard or expensive, and suggest changes that make a tight dimension easier to hit — so the tolerances on your print are ones the parts can actually meet. Send your part; NDA first.

Book a 30-minute call