Injection Molding Tolerances: What Is Actually Achievable

1. The myth of "the moulding tolerance"

Search for injection moulding tolerances and you will find tables that say things like "±0.05 mm for dimensions under 10 mm, ±0.10 mm for 10 – 100 mm". These tables are not wrong, but they are misleading in a specific way: they describe what a well-controlled process can achieve on a favourable feature, and they get copied into drawings as though they were a promise.

The honest statement is that tolerance is a function of five things at once — the material, the part geometry, the tool, the machine and the process control — and no single table can account for all of them. Two parts with the same nominal dimension can differ by a factor of five in achievable tolerance because one has a dimension across a parting line and the other has a dimension on a ground core pin.

2. Four sources of dimensional variation

Every deviation from nominal on a moulded part comes from one of four sources. Separating them is the whole of tolerance engineering, because each one is fixed differently.

Tool variation

The steel is machined to a nominal that already includes a shrinkage allowance. Tool variation is the sum of machining accuracy, EDM accuracy, fitting tolerance between cavity and core, and wear over the life of the tool. It is essentially fixed at build and then drifts slowly.

Typical tool-only variation on a well-built tool: ±0.02 – 0.05 mm on a machined dimension, ±0.01 mm on a ground dimension. This is the floor of what the process can ever achieve, because the process cannot correct for steel that is in the wrong place.

Shrinkage variation

Plastic shrinks as it cools, and the amount depends on the material, the wall thickness, the flow orientation, the packing pressure and the cooling rate. A tool cut for 0.5 % shrinkage will produce parts that are 0.45 % small on one dimension and 0.55 % on another, because the two dimensions did not see the same conditions. This is usually the largest single contributor on a new part.

Process variation

Cycle-to-cycle variation from melt temperature drift, hydraulic response, packing pressure variation, cooling time tolerance and ambient conditions. On a stable machine running a stable process this is small — a few hundredths of a millimetre. On a machine shared between jobs, with the process adjusted by whoever is on shift, it can be ten times larger.

Measurement variation

How the part is measured, at what temperature, with what instrument, by whom. A flexible part measured with a caliper under hand pressure reads differently from the same part on a CMM. A part measured straight off the machine, still warm, is larger than the same part measured the next day. This source is routinely ignored and it accounts for a surprising share of "the parts are out of tolerance" conversations.

Shrinkage scales with the dimension; the other three are closer to absolute. That is why large parts need percentage-based tolerances, not fixed ones.

3. Shrinkage: the number that is never one number

Every material datasheet gives a shrinkage value, usually a range like 1.4 – 2.0 %. What the datasheet does not say is that the value depends on wall thickness and flow direction, and that a tool designed to the midpoint of the range will be wrong on some dimensions.

Directional difference means the part shrinks more across the flow direction than along it. On a long, flow-oriented part this alone can consume the whole tolerance.

Three practical consequences follow.

Shrinkage is not linear with wall thickness. A 2 mm wall and a 3 mm wall in the same material do not shrink by the same percentage — the thicker section shrinks more, because it stays hot longer and has more material to contract. A part with mixed wall thickness will have dimensions that drift in different directions.

Glass-filled materials are anisotropic. Glass fibres align with the flow, and the resin shrinks much less along the fibres than across them. A 30 % glass-filled PA66 part can shrink 0.3 % along the flow and 0.8 % across it. If the tool is cut to a single shrinkage value, one direction will be out of tolerance no matter what the process does.

Polyamides keep moving after they leave the machine. PA6 and PA66 absorb moisture from the air, and a dry-as-moulded part grows by 0.3 – 0.5 % over the first weeks until it reaches equilibrium. A part that measures in tolerance on the day it is moulded can be out of tolerance after a month in a humid warehouse. For precision PA parts, the drawing has to state the condition — dry as moulded, or conditioned.

4. Tolerance grades and how to use them

The two systems you will meet are DIN 16742 (now largely superseded by ISO 20457) and the supplier-specific tables that grew out of them. Both work the same way: a tolerance grade is selected from a table based on the nominal dimension and the material group, and the grade accounts for the fact that shrinkage variation scales with size.

Indicative values for unfilled commodity and engineering resins, single-cavity tool, controlled process. Glass-filled and semi-crystalline resins move one grade; dimensions across a parting line or on a slide move one grade.

Grade 2 is the sensible default for a general mechanical part. Grade 1 is for dimensions that are functionally critical and is a deliberate commercial decision — it usually implies a single-cavity tool, a dedicated machine, SPC and a first-article inspection report. Grade 3 is for dimensions nobody will ever measure, and specifying it honestly saves money.

The trap is that most drawings use a single general tolerance block for everything, so a dimension that could be ±0.5 mm is specified at ±0.1 mm along with the one dimension that genuinely needs it. That is how a straightforward part gets quoted as a precision part.

5. Achievable tolerance by feature type

The feature type matters more than the dimension. Two dimensions of the same nominal size can differ by a factor of three in achievable tolerance depending on how the steel that forms them is made and how the part releases from it.

Note how flatness and warpage sit outside the pattern. They are not tool tolerances at all — they are process and design outcomes, and no amount of tool precision fixes them.

The parting line row deserves emphasis because it is the most common source of unplanned variation. A dimension that crosses the parting line is formed by two steel blocks that have to meet. The closing tolerance, the wear on the leader pins, the clamp force distribution and the thermal expansion of each half all contribute. If a critical dimension crosses the parting line, consider redesigning so it does not.

6. Tolerance by material

Material affects tolerance through shrinkage magnitude, shrinkage anisotropy, moisture absorption and stiffness. A flexible part is also harder to measure, which is a tolerance problem in its own right.

The modifier stacks with the feature modifier. A glass-filled PA66 dimension across a parting line is a grade 3 dimension, not a grade 1 dimension, no matter how it is drawn.

For elastomers the measurement problem is as significant as the shrinkage. A TPE part measured with a caliper compresses under the jaw, and two inspectors will get different numbers from the same part. TPE tolerances should be agreed with a specified measuring method — usually a non-contact optical measurement or a fixture that controls the compression.

7. Datums, and why most disputes start here

A tolerance is meaningless without a datum. If a drawing says a hole must be within ±0.1 mm of something, that something has to be a feature that exists on the part, is stable, and is accessible to the inspector.

The failure mode is common: the drawing dimensions everything from a theoretical centre or from a feature that is itself toleranced, so the inspector has to choose a reference, and their choice differs from the designer's assumption. The part then "fails" inspection even though nothing is wrong with it.

• Choose mouldable datums. A flat face, a moulded boss, a machined-in step. Not a centreline through a curved surface, and not a feature that warps.
• State the primary, secondary and tertiary datum explicitly. A three-plane datum scheme removes almost all ambiguity.
• Put the datum on the functional surface. If the part locates on a specific face in the assembly, that face is the datum — not the opposite face that happens to be flatter.
• Watch out for datums on warped features. If the primary datum is a large flat panel that warps 0.5 mm, every dimension measured from it inherits that variation.

8. What Cpk actually tells you

A first-article inspection report tells you whether the parts you measured were in tolerance. It does not tell you whether the process will stay in tolerance. That is what Cpk measures, and it is the number worth asking for on a critical dimension.

Cpk compares the tolerance width to the process spread, centred on the nominal. A Cpk of 1.33 means the process spread is about three-quarters of the tolerance width and the process is centred — roughly 60 parts per million outside the limits if the distribution is normal. A Cpk of 1.0 means the spread equals the tolerance, and you should expect about 2,700 ppm out of tolerance.

Values assume a normal distribution and a centred process. Moulding processes are often not perfectly normal, so treat the defect rates as indicative rather than exact.

For most moulded parts, Cpk 1.33 on the two or three genuinely critical dimensions is the right target. Demanding Cpk 1.67 on everything is a way of paying for process control you do not need.

9. What tight tolerances cost

Tolerance is not free, and the cost is not linear. Halving a tolerance typically doubles or triples the cost of holding it, because each step requires a different class of tool, machine and control.

Cost impact includes tool, piece price and inspection. On a high-volume part the inspection cost alone can exceed the tool cost difference.

Multi-cavity tools add a specific problem: cavities are not identical. Two cavities cut to the same nominal can differ by 0.02 – 0.05 mm because of machining and fitting, and they can shrink differently because their cooling circuits are not thermally identical. On a ±0.05 mm dimension, an eight-cavity tool is eight separate processes that all have to be capable. This is why precision parts are usually single-cavity even when the volume justifies more.

10. Measuring a moulded part properly

A large share of tolerance disputes are measurement disputes. Four practices remove most of them.

Measure at a defined time and condition. Plastic parts continue to shrink and, for polyamides, to absorb moisture for hours or days after moulding. State whether the measurement is at 1 hour, 24 hours or 48 hours, and at what temperature and humidity. For PA parts, state whether the part is dry or conditioned.

Use a fixture for flexible parts. A caliper on a TPE or thin-wall part reads whatever the operator's hand pressure produces. A simple gauge that locates on the datum features gives repeatable numbers; a caliper does not.

Match the instrument to the tolerance. A caliper resolves 0.01 mm but its practical accuracy on a plastic part is closer to ±0.05 mm. Measuring a ±0.05 mm tolerance with a caliper is measuring noise. Use a micrometer, a height gauge on a granite plate, or a CMM.

Agree the measurement method in writing before the first article. Which dimensions, which instrument, which datums, at what time after moulding, and who signs off. A first-article report without an agreed method is an opinion.

11. How to write a tolerance block that works

Most tolerance problems are created on the drawing, before anyone touches steel. A tolerance block that works has four properties.

• One general tolerance, honestly chosen. Put the realistic general tolerance in the title block — usually grade 2 or grade 3 — and then call out the handful of dimensions that need better. This is the opposite of the common practice, which is to put ±0.05 mm in the title block and then argue about it.
• Functional tolerances only. Tolerance a dimension because something fits, seals, moves or is measured by the customer. If nothing depends on it, it does not need a tolerance.
• Stated datums. Three planes, explicitly marked, on features that exist and do not warp.
• Stated condition. Material, moisture condition, time after moulding, and the measurement standard for anything flexible.

A useful discipline is to mark each toleranced dimension with the reason it is toleranced — "seal groove", "bearing seat", "mates with part 2". When the toolmaker asks why a dimension is ±0.05 mm, the answer is in the drawing rather than in someone's memory, and when the part fails inspection everyone can see what actually matters.

12. Frequently asked questions