Tolerance Stack-Up in Multi-Part Assemblies — How Do You Account for Mold Shrinkage, Process Variation and Assembly Fit When Designing Interlocking Injection Molded Parts?

Last updated: 2026-09-03 · Reading time: about 6 min

Tolerance stack-up is the cumulative variation across the mating parts of an assembly — each part varies within its drawing tolerance, and the variations combine along the chain. Run it two ways: worst-case sums every tolerance (conservative, for safety-critical fits); RSS takes the root sum square (about half the worst-case, realistic for chains of 5+ parts). The assembly fits only if the stack stays inside the clearance.
See also related article.

Key facts

<>

  • Two methods, two philosophies: worst-case (arithmetic sum) assumes every part peaks at its tolerance extreme at once — use it for safety-critical fits and short chains of 2–3 parts; RSS (root sum square) assumes a normal spread, totals about half the worst-case, and suits chains of 5+ parts where 99.73% of assemblies fit at ±3σ.
  • Six sources feed the stack: mold shrinkage (±0.05–0.15 mm/100 mm), machining (±0.01–0.05 mm CNC, ±0.005–0.02 mm EDM/grinding), shot-to-shot process (±0.02–0.10 mm), resin batch MFI (±0.03–0.08 mm/100 mm), environment (±0.01–0.05 mm/100 mm for PA, PC) and mold wear (±0.01–0.05 mm over 100,000 shots).
  • The fit test: if the stack exceeds the nominal clearance the parts bind or crack (interference); if it is large relative to the clearance they rattle or leak (excess play). Confirm the assumed process variation against 20–30 T1 samples measured by CMM before the tolerances are locked.

    • Parting line — The line where the two mold halves (A-side and B-side) meet. Parting line location affects: gate placement, ejector pin marks, witness lines, and cosmetic appearance. A well-designed parting line minimizes visible witness marks. See also related guide.

    What are the sources of variation — and how much does each contribute?

    Six sources, each with its typical magnitude and the lever that reduces it.

    Variation source Typical magnitude How to reduce it
    Mold shrinkage variation ±0.05–0.15 mm/100 mm (resin-dependent) Use the supplier’s shrink range, not one value; design the cavity to the range midpoint
    Mold machining tolerance ±0.01–0.05 mm CNC; ±0.005–0.02 mm EDM/grinding CNC for rough features; EDM or grinding for critical ones
    Process variation (shot-to-shot) ±0.02–0.10 mm Cavity-pressure monitoring, stable melt and mold temperature
    Resin batch variation (MFI) ±0.03–0.08 mm/100 mm (MFI ±15% between batches) One batch per run; re-tune the process for each batch
    Environmental (temperature, humidity) ±0.01–0.05 mm/100 mm (hygroscopic PA, PC) Condition parts 24 h at 23°C, 50% RH before assembly
    Mold wear (over life) ±0.01–0.05 mm over 100,000 shots (abrasive GF resin) Hardened steel (H13, 48–52 HRC) on critical features; monitor dimensions periodically

    Combine the six by root sum square, not by addition: total = √(shrinkage² + machining² + process² + batch² + environment² + wear²). For a 100 mm ABS dimension that is √(0.10² + 0.03² + 0.05² + 0.05² + 0.02² + 0.02²) = √0.0167 ≈ ±0.13 mm.

    The drawing tolerance must sit wider than that process capability — ±0.15 or ±0.20 mm here — or parts fall out of tolerance no matter how well the mold is built. Set the drawing tighter than what the process can hold and you are not specifying precision, you are specifying scrap.

    How do you run the stack-up — and which method fits the assembly?

    Four steps, from the tolerance loop to the T1 check.

    1. Draw the tolerance loop. Trace the chain from one datum on the assembly to the mating feature and list every part dimension that adds to the gap — wall thicknesses, pin and bore diameters, datums and the parting-line position. A dimension belongs in the chain only if changing it changes the fit; anything off the loop is ignored.
    2. Choose worst-case or RSS. Worst-case adds the tolerances arithmetically — 4 parts at ±0.10 mm give ±0.40 mm — assuming all of them peak at once; it is conservative and right for safety-critical fits and short chains of 2–3 parts. RSS takes the root sum square — the same 4 parts give √(4×0.10²) = ±0.20 mm — realistic for 5+ parts, where 99.73% of assemblies fit at ±3σ.
    3. Budget the clearance across the chain. The nominal gap is the total you can spend. For a snap-fit lid at 0.15 mm clearance with a 0.05 mm safety margin, the stack must stay under 0.10 mm; split evenly across 4 dimensions that is ±0.025 mm each. If ±0.025 mm is tighter than the process holds (±0.05 mm), loosen the tolerance and widen the clearance to 0.25 mm rather than specify a capability the mold cannot .
    4. Verify with T1 samples. The stack assumes a process variation you have not measured yet. Mold 20–30 shots, CMM every critical dimension and compute the real σ. If actual variation exceeds the assumption, re-run the stack and adjust; if it comes in lower, the tolerances can open up and the mold cost drops.

    A stack-up is a budget, and the assembly clearance is the total you can spend. Every part in the chain draws from it. Worst-case assumes they all max out at the same instant; RSS assumes a normal spread where they rarely do. Choose the method by how critical the fit is and how many parts feed the loop — then confirm the real variation at T1.

    “The stack-up is where an assembly either fits on paper or fails at T1. We build the tolerance loop with the customer before cutting steel — worst-case for a safety-critical snap, RSS for a long chain — and give each dimension its share of the clearance. Then we measure 20–30 T1 samples and re-run the numbers with the real σ. If the process can’t hold what the fit needs, we widen the clearance, not the scrap rate.”

    — Project Engineer, aCatMould

    Frequently asked questions

  • Key parameter identified — critical variable confirmed against spec.
  • Tolerance within achievable range — commercial ±0.15mm or tighter as needed.
  • Material matched to application — resin grade selected for end-use environment.
  • Process window validated at T1 — parameters locked after first trial.
  • Inspection method defined — CMM, visual, or functional test specified.

    • How do you handle two mating parts molded from different resins?

    Different resins shrink at different rates — PP 1.5–2.0%, ABS 0.4–0.7%, PC 0.5–0.7% — so the mating dimensions move by different amounts and the clearance drifts.
    A PP bore and an ABS pin at a 0.20 mm nominal gap can open to 0.40 mm once each shrinks against its own mold dimension.

    The fix sits in the mold dimensions, not the drawing: size each cavity so the shrunken part lands where the fit needs it. The designer takes the supplier’s shrink-range midpoint for the nominal mold size and feeds the range extremes into the stack as the tolerance input.

    What is the minimum clearance for a snap-fit — and how do you stop it cracking?

    For a cantilever snap the clearance is set by the deflection the beam needs to engage the hook, and the ceiling is the resin strain. Maximum strain ≈ 1.5·t·y/L² (t = beam thickness, y = deflection, L = beam length); it must stay under the yield strain — about 2–3% for ABS, 6–7% for PC, 8–12% for PP.

    A 15 mm PP beam 1.5 mm thick deflects about 1.0 mm before it yields, so the hook needs 1.0 mm of clearance. The stack eats into it: ±0.20 mm of variation drops the minimum to 0.80 mm — less retention — and pushes the maximum to 1.20 mm, where beam stress rises. Check both ends.

    Does GD&T help the stack-up?

    Yes. GD&T (ASME Y14.5 / ISO GPS) specifies form, orientation, location and runout against a datum reference frame, so every part is measured — and every stack is calculated — from the same reference. That alone removes a common source of assembly error.

    It also unlocks bonus tolerance: a position tolerance at maximum material condition grows as the feature departs from MMC, adding 0.01–0.05 mm of slack that loosens the stack and cuts scrap. The stack is then run on the virtual condition — the worst-case envelope of size plus geometry — not on size tolerance alone.

    Key takeaways

    1. A stack-up is a clearance budget: the nominal gap is all you can spend, and every dimension in the tolerance loop draws from it. Trace the loop first — only dimensions that change the fit belong in it.
    2. Pick the method by criticality and chain length — worst-case (arithmetic sum) for safety-critical fits and 2–3 parts, RSS (root sum square, about half the worst-case) for 5+ parts — then give each dimension its share of the clearance.
    3. The stack is only as good as the process σ you assume. Verify it with 20–30 T1 samples measured by CMM; if the process can’t hold the tolerance the fit needs, widen the clearance rather than specify scrap.

    Tolerance stack-up turns a fit from a hope into a calculation. Build the loop, choose worst-case or RSS by how critical the assembly is, budget the clearance across every contributing dimension, and confirm the real variation at T1. Do it before the mold is cut — skip it and you find the stack at the worst moment, when the parts are already molded.

    Sources
    ISO 20457:2018 — Plastics moulded parts: Tolerances and acceptance conditions
    Mitutoyo — CMM and measurement instruments for dimensional verification
    MatWeb — material property database for resin shrinkage and modulus
    Plastics Technology — mold design, shrinkage and dimensional control practice
    MoldMaking Technology — mold build tolerance and process capability

    Sources: Plastics Technology · Plastics Technology Online · Society of Plastics Engineers (SPE) · ThomasNet