Last updated: 2026-09-02 · 9 min read

What Is Mold Trial Repeatability — and How Many Shots Prove It?

Repeatability is the tool-plus-process producing the same part shot after shot without adjustment — and it is proven by a defined, unadjusted run of consecutive shots with weight and dimensional records, not by a handful of good parts shown after the fact. The honest answer to “how many shots” is tiered by purpose: a function check needs tens, a process validation needs a hundred or more, and a production handover needs a continuous run long enough to expose drift.

Ejector pin — A steel pin that pushes the molded part out of the cavity after cooling. Typical diameter: 2–10mm. Pin marks are visible on the part surface — position at non-cosmetic areas. Pin count and placement affect ejection force and part warpage. See also related guide.

Gate vestige — The visible mark left on the part surface where the gate was trimmed. Edge gate: 0.5–2.0mm vestige. Pin-point gate: 0.2–0.5mm. Valve gate: 0.05–0.1mm (flush). Vestige size affects cosmetic appearance and may require secondary trimming. See also related guide.

Key Facts

  • One good shot is an anecdote. Shot-to-shot variation only becomes visible across a consecutive series: with roughly 30 or more consecutive shots you can compute a meaningful standard deviation for part weight and key dimensions — enough to tell stable variation from a trend. Repeatability is about the width of the band, not the height of the average.
  • Each trial stage asks a different question, so each has its own shot count: T1 asks “does the tool work at all” (short run, 20–50 shots); the process validation run asks “is the process stable” (100–300 shots, unadjusted); the production handover asks “can it run an hour unattended” (a continuous timed run under monitoring, no operator touches).
  • Adjustments disqualify. A repeatability run with parameter changes in the middle proves nothing — the variation you measure is the variation of three different processes. The counted run starts when the process is frozen, and any change restarts it.
  • Warm-up comes before the count. Mold steel expands to thermal equilibrium over the first dozens of shots; dimensions and weight drift during that window even on a perfect tool. Stabilization shots run first, and the counted run begins only after mold and machine temperatures plateau.
  • Repeatability is the raw material of capability. The data from an honest run feeds the capability indices your quality team computes — which is why the run must record every shot’s weight and preserve numbered samples, not just a photo of five nice parts.

    • “The most expensive sentence in tooling is ‘the first shots looked great.’ Looked great at shot number three tells you nothing about shot number three hundred — I only sign acceptance after the counted run, and I restart it if anyone touches a dial.”
    — Arthur, Tooling Engineer at aCatMould

    What “Repeatable” Means in Numbers

    Repeatability is expressed as variation around a stable average: shot weight within a narrow band, critical dimensions within their tolerance with room to spare, and appearance stable across cavities. In practice you plot the run — weight per shot, dimension samples at set intervals — and read two things separately: the width of the band (random variation) and the direction of the middle (drift).
    A wide band means the process is noisy; a sloping middle means something is changing — mold temperature still climbing, resin drying drifting, or a check ring wearing. Both must pass judgment: a stable average with wild swings is not repeatable, and a tight band that slides hour by hour is not either.

    How Many Shots at Each Stage

    Stage Typical Count Question Answered Recorded
    T1 function run 20–50 shots Does the tool fill, pack, eject, and cool? Defects by cavity, weight, first dimension check
    Process validation run 100–300 shots Is the frozen process stable enough for capability? Weight per shot, dimension samples, curve data
    Production handover 1–2 h continuous Can it run unattended without drift? Monitoring alarms, weight trend, downtime events
    Resin lot change check 30–50 shots Does the new lot sit in the same window? Weight overlay vs baseline, first-off dimensions

    The Warm-Up Problem Nobody Counts

    A mold reaches thermal equilibrium later than people expect: the steel expands, the cooling circuits settle, and the part dimensions drift through the first stretch of running even when everything is correct. A counted run that starts at shot one mixes that transient into the statistics and poisons the conclusion.
    The discipline is to run a warm-up phase — often 20 to 50 shots depending on mold mass and cycle time — confirm the mold temperature readings have plateaued, freeze the process, and only then start the counted run. The warm-up shots are not wasted; they are the price of a meaningful number.
    Record the plateau point itself — the shot number where temperatures settled — because it is a useful machine-and-mold pairing fact for every future setup of this tool.

    Rules That Keep the Run Honest

  • Freeze the process before shot zero of the counted run — any adjustment restarts the count, and the log records why.
  • Weigh every shot of the run (or every shot of a defined sampling pattern), not just the first and last; the trend lives in the middle.
  • Preserve numbered samples at set intervals — for example shots 10, 25, 50, 100 — bagged and labeled by shot number and cavity, so the dimension record maps to the run.
  • Track every cavity separately; a per-mold average hides the one cavity that will become your problem in production.
  • Define drift alarms before the run starts (weight band, curve windows) so “stop and investigate” is a written rule, not a judgment call in the moment.
  • Record the environment too: water temperatures, resin drying readout, and hopper state — the variables that explain the drift if one appears.

    • Reading the Results: Variation vs Drift

    1

    Plot weight against shot number

    A flat band of points means the process holds; a steady slope means drift — most often thermal (mold not fully stabilized) or drying (moisture creeping into a hygroscopic resin).
    A step change means an event: someone changed a setting, a lot was topped up, or a water line was disturbed.

    2

    Overlay the dimension samples

    Check the numbered samples against the drawing at the intervals you bagged them. Dimensions that track the weight trend share a cause with it; dimensions that wander on their own point at a specific feature — a core shifting, a gate wearing, an ejector stroke changing.

    3

    Separate cavities before judging

    Per-cavity plots answer the question a merged chart cannot: is the whole mold noisy, or is one cavity the outlier? A single outlier is a tooling conversation; a uniform band is a process conversation. Confusing the two is how shops “fix” a worn gate by changing the drying protocol.

    4

    Fail means restart, not resume

    If the run fails its windows, the correction happens, the mold re-stabilizes, and the counted run restarts from zero. Partial credit is how marginal tools slip into production and write their defect history on your floor instead of at the trial.

    What to Hand Over With the Run

  • The frozen setup sheet — every parameter, so the run is reproducible on the next machine without archaeology.
  • The weight record per shot with the band and any alarms that fired, kept with the run’s time stamps.
  • The numbered dimension samples with their measurement report — the same samples, not re-molded substitutes.
  • The per-cavity map from the run, so future drift has a baseline to be measured against.
  • The process monitor data (curves, alarm log) if the tool runs instrumented — the continuous witness that goes home with you.

    • Frequently Asked Questions

    Is 100 shots always enough?

    It is a reasonable floor for a stable-process judgment on most tools, but two modifiers apply.
    Fast-cycling small parts can and should run longer in wall-clock time — several hundred shots in under an hour — to also prove the unattended-run question. And safety-critical or tight-tolerance parts justify a longer counted run, because the cost of a wrong conclusion dwarfs the machine time saved.

    Can I shorten the run for a simple bracket?

    Yes — tier the count to risk. A non-critical bracket with generous tolerances may pass its handover on a shorter defined run with weight tracking and a first-off dimension check. What you may not shorten is the discipline: the run is unadjusted, counted, and recorded, whatever its length.

    Why did dimensions drift during an otherwise clean run?

    The two usual suspects are thermal and moisture. Thermal: the mold had not fully plateaued when the count started, so steel growth carried the dimensions. Moisture: a hygroscopic resin drying below its window shifts viscosity and shrinkage over the run. The weight plot distinguishes them — drying drift usually shows in weight too; pure thermal growth often does not.

    Do all cavities really need tracking on a 32-cavity tool?

    Weight can be sampled by a defined pattern as long as every cavity is covered across the run; dimensions get sampled per cavity position on a schedule. What matters is that no cavity is invisible: the outlier cavity is the one that generates your first customer complaint, and the run is your only chance to see it before shipping.

    What if the process was changed mid-run for a good reason?

    Then the run stopped being a repeatability run at that shot. Record the change, let the mold re-stabilize, and restart the counted run. The time is not lost — the earlier segment documents why the change was needed — but no acceptance signature rides on a run with a dial turn in its middle.

    How does repeatability relate to capability indices?

    The run produces the data; capability indices are computed from it against the drawing’s tolerance. A repeatable run with tight variation yields meaningful indices; a noisy or drifting run yields numbers that mislead. That is why the counted run comes first and the indices second — computing capability from an adjusted, partial run is statistical theater.

    Who owns the counted-run definition — buyer or molder?

    It is agreed, not owned — and it is written into the acceptance criteria before T1. The molder brings the machine reality (cycle time, shot count per hour, what a continuous run costs); you bring the risk view (which parts justify a longer run, which tolerances need the samples numbered). A definition agreed after the run is a negotiation; one agreed before it is a specification.

    What if the tool repeats at the molder but not in my plant?

    Compare the two environments line by line: machine size and screw design, water temperature control quality, resin drying, and the setup sheet as actually entered. The usual culprits are drying discipline and water temperature stability — variables the trial floor controlled tightly and the production floor assumes. The handover documents exist precisely to make this comparison a checklist instead of a dispute.

    Key Takeaways

    1. Repeatability is proven by a consecutive, unadjusted, recorded run — tiered by purpose: tens of shots to prove function, a hundred-plus to prove stability, a continuous timed run to prove the handover.
    2. Warm-up precedes the count, adjustments restart the count, and every cavity stays visible throughout — these three rules keep the result honest.
    3. Read variation and drift separately on the weight plot, keep numbered samples mapped to shot numbers, and fail means restart — partial credit is how marginal tools production.
    4. Hand over the setup sheet, weight record, samples, per-cavity map and monitor data together; the capability indices your quality team computes are only as honest as that record.
    5. Define the run’s length and windows in the acceptance criteria before T1, so “how many shots prove it” is a number you agreed on, not a debate after the fact.

    Repeatability is where tooling trust is either earned or postponed. aCatMould writes the counted-run definition into the acceptance criteria before steel ships — stage by stage, with warm-up, alarm windows and per-cavity records specified — so the tool arrives with its stability already documented, and your first production week looks like the trial, not like a second trial. The counted run costs machine hours; the argument it prevents costs calendar months. Between those two prices, the machine hours are the bargain every time.

    Sources
    Plastics Technology — process validation and monitoring
    MoldMaking Technology — mold trial best practice
    Plastics News — tooling acceptance practice

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