Why Do Part Dimensions Drift Between Resin Lots?

Last updated: 2026-09-01 · Reading time: about 10 min

The same mold, the same settings, a new pallet of resin — and the parts measure a little different. Lot-to-lot dimensional drift is normal physics meeting normal supply: resins vary slightly between batches in flow and additives, processes drift slowly in temperature and timing, and the part’s size is the sum of both.
The discipline is not eliminating drift — it is knowing which input moved, keeping the drift inside a shrinkage window that the tolerances were designed to absorb, and catching the exceptions with first-article checks before they cost a batch. This article maps the moving parts and the controls.

Key facts

<>

  • A dimension is not a point, it is a cloud: molded size is set by a shrinkage window, not a constant — and tolerance design that assumes a single number will fight every new lot, as the tolerance page explains.
  • Resin-side movers concentrate in a short list: melt-flow index, moisture, regrind share and colorant batch — each measurable at goods-in, each with a known direction of effect.
  • Process-side movers are slower but steadier: melt and mold temperature, hold pressure and time, cycle rhythm — and a machine change moves the whole distribution at once.
  • The first-article check on every new lot is the control that pays for all the others: it converts drift from a complaint at assembly into a decision at the press.
  • Records decide verdicts: lots, settings and measurements kept together close a drift case in hours; kept apart, the same case runs for weeks.

    • Cavity — The hollow space inside the mold that forms the part shape. Multi-cavity molds produce multiple parts per shot. Cavity count determines production output but also mold cost and complexity. Each cavity must be balanced for uniform fill. See also related guide.

    What drifts between lots — and why does size move?

    Lot-to-lot variation
    The batch-to-batch differences within a resin’s specification — melt-flow index band, additive dosing, moisture pickup, colorant carrier — plus the slow process drift between production runs, seen together as small shifts in part dimensions.

    The mechanism runs through shrinkage. Molded size is steel size minus shrinkage, and shrinkage is not a resin constant — it is a response to the material’s state and the process’s hand: the shrinkage page maps the levers. A new lot arrives with its melt-flow index somewhere inside the datasheet band — a higher number flows easier, packs differently and shrinks slightly differently; the MFI page shows the sensitivity.
    Moisture changes the equation twice — as hydrolysis risk and as a packing disturbance — which is why the drying record is part of every dimensional story. Colorant lots shift the carrier ratio; regrind percentages wobble with scrap rates. None of these inputs is out of specification — that is the point: they are all inside their bands, and the part size wanders within the window those bands define.
    The tolerance conversation at drawing time decides whether that window fits inside the drawing’s callouts; the drift conversation at production time decides whether this week’s inputs are behaving like last week’s.

    Which input moved?

    Observation Points to Confirming check
    All dimensions shifted the same direction after a delivery New resin lot — MFI or additive band moved. Compare lot MFI against the previous lot’s certificate.
    Size moved with surface splay or streaks Moisture — drying regime changed or failed. Read the drying log; run a dried comparison.
    Shift appeared with a new colorant delivery Colorant carrier batch changed the blend. Hold colorant constant for a run and compare.
    Gradual drift over days, same direction Process creep — temperature, hold pressure or cycle rhythm. Compare current settings against the documented window.
    Step change after maintenance or a press swap Machine characteristics — plastication or valve behavior. Run the documented window and re-baseline if needed.
    Measurements moved, parts look identical Measurement system — gage, fixture or operator change. Re-measure retained parts on both setups.
    Scatter widened, center stayed Instability, not shift — cycling inputs, worn valve, loose fitting. Trend chart on key dimensions over the last runs.

    The table works because the suspects leave different signatures: lot changes move everything at once, process creep moves slowly in one direction, machine changes step the whole distribution, and measurement issues move the numbers without touching the parts.
    Reading the signature first saves a week of shotgun debugging. It also disciplines the conversation with suppliers: a claim that “the resin changed” carries weight only with certificates beside it, and a claim that “nothing changed on the floor” carries weight only with the settings log beside it. Both documents cost minutes to keep and decide the case.

    How do you catch drift before it costs parts?

    1Gate every new lot with a first-article check. A handful of parts measured on the critical dimensions before the lot enters production — the single control that catches resin-side drift at the press instead of at assembly.
    The acceptance discipline already proves the tool; this proves the material.
    2Keep the parameters in a documented window. Melt and mold temperatures, hold pressure and time, cycle time — written as the window that produced good parts, so any drift is a departure from a record, not from memory.
    3Trend the critical dimensions. A simple chart of measured size over runs turns slow creep into a visible slope — the drift you can see Tuesday is the scrap you avoid on Friday.
    4Compare certificates at goods-in. MFI, moisture rating and colorant lot read against the previous delivery — five minutes at receiving that explains most Monday surprises.
    5Pre-agree the response rules. What triggers a hold, who authorizes a window adjustment, when the supplier is called — decided before the deviation, so the response is a procedure instead of a debate.

    How should tolerances absorb normal drift?

  • Callouts sized to the shrinkage window — critical dimensions given tolerances the material’s window plus process spread can hold, per the realistic ranges.
  • Critical dimensions named, not assumed — the drawing declares which few dimensions carry the function; everything else takes standard tolerance.
  • Tool tolerance separated from process tolerance — steel cuts set the center; the process sets the spread; each is audited by its own evidence.
  • First articles triggered by change — new lot, new machine, new shift after maintenance — the checks fire on change, not on calendar.
  • Measurement fixed before parts are judged — same gage, same fixture, same temperature conditions; a drifting gage manufactures phantom drift.
  • Records kept as one file — lot certificates, settings, measurements archived together, so the next investigation starts from evidence.

    • What does a drift investigation look like?

    Fast ones share a shape: the signature read first — everything moved at once (lot), slowly in one direction (creep), stepped (machine), or only the numbers moved (measurement); then the paired documents pulled — certificates against certificates, settings against the documented window, and retained parts re-measured to clear the gage.
    Filled resins add a wrinkle worth naming: their shrinkage windows are narrower, which helps the absolute numbers, but the filled-grade sensitivities mean fiber-related inputs sit in the suspect list too. Most cases end at the certificate or the settings log; the ones that go deeper end at a controlled comparison run — one input changed at a time, measured on the same gage, judged against the trend chart.
    What investigations should never do is adjust three settings in one afternoon and declare victory: that hides the real input until the next lot, at a higher price. And when the verdict lands on the resin band itself — the tolerance genuinely cannot absorb the specification’s width — the conversation moves to the drawing or to a tighter material spec, both legitimate, both less painful the earlier they are raised.
    Either conversation is a design-stage instinct: the program that audits its shrinkage window at drawing time meets every new lot with a shrug instead of an investigation, and the program that skips the audit meets the same physics holding a rejected batch.

    “Drift is rarely one bad input — it is a signature. Everything moving at once is the resin; creeping one way is the process; stepping overnight is the machine. Read the signature and the suspect confesses.”
    — Arthur, Tooling Engineer at aCatMould

    Frequently asked questions

    Is lot-to-lot drift the supplier’s fault?

    Not usually, and treating it as blame wastes the investigation. Resins ship inside specification bands, and bands have width — that is the industry’s normal. The supplier’s obligation is consistency of data: certificates that state what actually shipped, so the molder can respond.
    The buyer’s protection is the tolerance design and the first-article gate, both of which absorb or catch the normal width. Fault enters only when certificates lie or shipments fall outside the stated band — and both are proven by records, not by suspicion.

    Can we buy tighter resin specifications?

    Often yes — a narrower MFI band, dedicated colorant lots, moisture-protected packaging are all purchasable, and some programs genuinely need them. The price rises with the tightness, which is exactly why the tolerance analysis comes first: if the drawing can absorb the standard band, the extra cost buys nothing; if it cannot, the tighter spec is the correct engineering purchase. Raise it at the datasheet conversation, not after the first rejected batch.

    Why did dimensions move after changing machines?

    Every press plasticates differently: screw geometry, check-valve behavior, barrel heating and shot-to-shot consistency each leave a fingerprint on melt state, and melt state is a shrinkage input. Two machines running identical settings are not identical processes — the settings are a shared vocabulary, not a shared machine. The practical answer is re-baselining: run the documented window, measure first articles, and adjust into the window that reproduces the qualified part — then document that machine’s window as its own.

    Do filled resins drift more than unfilled ones?

    Differently. Their shrinkage is lower and the absolute numbers sit closer to steel size — helpful — but the fiber introduces its own batch variables: fiber length distribution and dispersion shift the anisotropy slightly lot to lot, and the warp-sensitive geometries feel it first. Filled programs benefit most from the first-article gate and from trending flatness alongside length dimensions.

    How much drift is normal?

    Enough to matter and too variable to quote as a universal number — the honest frame is the window: drift is normal while dimensions stay inside the band defined by the resin’s shrinkage window plus the documented process spread. What is never normal is movement outside that band, scatter that keeps widening, or a direction that keeps going after two lots. Those three are investigation triggers, and they are worth naming in the response rules before anyone needs them.

    Who owns the fix — resin, process or mold?

    The signature assigns it. Lot-side causes belong to material management — specification, goods-in checks, storage and drying. Creep belongs to process control — windows, trend charts, response rules. A mold that produces parts whose scatter keeps growing belongs on the maintenance bench — wear shows up in dimensions before it shows in appearance. The rare case where the drawing itself cannot live with normal physics goes back to the tolerance conversation, and that is a design decision, not a production failure.

    Key takeaways

    1. Dimensional drift between lots is normal physics: resin bands plus process spread define a shrinkage window, and size wanders inside it.
    2. The signature tells the suspect: lot shifts move everything, creep moves slowly, machines step, measurement moves numbers without parts.
    3. The first-article check on every new lot is the highest-value control; certificates at goods-in explain most surprises in five minutes.
    4. Tolerances designed to the shrinkage window absorb normal drift; those that cannot are a drawing conversation, raised early and cheaply.
    5. Records — certificates, settings, measurements kept together — close drift cases in hours instead of weeks.

    If dimensions are wandering between lots, send the drawings’ critical callouts, the last two lot certificates and your current settings. You will get the signature read and the controls ranked — from a named tooling engineer, before the next delivery lands. The records that decide the case take minutes to keep, and they turn every future lot into a routine check instead of a debate.

    Sources
    ISO 20457:2018 — Plastics moulded parts: Tolerances and acceptance conditions, International Organization for Standardization
    MatWeb — material property database, reference for specification bands and shrinkage data
    Plastics Industry Association — industry resources on molding practice
    Autodesk Moldflow — injection molding simulation, reference for shrinkage sensitivity analysis

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