A mold trial is not a product viewing. It is the only point in the project where the tool, the process and the measurement system are all on the table at once, and it is the last moment when fixing a problem costs a few days instead of a rework cycle. Most buyers treat it as a courtesy call and sign whatever is put in front of them. This guide covers what to measure, when to measure it, and the five conditions that should stop a sign-off.
What a mold trial is actually for
The usual scene: a box of samples arrives, someone opens it, turns a part over twice, says it looks good, and a sign-off email goes out the same afternoon. Two months later the parts are being rejected at incoming inspection, or the tool cannot hold tolerance on a different machine, or the cosmetic surface that looked fine in the sample looks different under the customer’s lighting.
None of that is bad luck. It is the predictable result of treating a trial as a look rather than as a test.
A trial exists to answer four questions, and a sample box can only answer the first one. The other three all have to be asked of the supplier while the tool is still on the machine, because once it ships, every answer becomes a plane ticket and a maintenance window.
It is also the point where the trial rounds stop being a line item and start being a cost. If you want to see where sampling sits inside a tooling quotation, the breakdown is in what drives the cost of an injection mold.
T0, T1, T2 — agree on the words first
Before anything else, pin down terminology. The injection molding industry uses T-numbers loosely, and the definitions shift between suppliers, regions and customers. A quotation that promises "two trials included" is worth very different things depending on whether the supplier counts T0 as a trial.
| Name | What it is | What you should receive | Who is normally present |
|---|---|---|---|
| T0 | First closing of the tool on a machine, usually internal | Nothing, or a verbal report | Toolmakers only |
| T1 / FOT | First formal sampling, parts sent for your evaluation | Sample set, trial report, dimensional report | Toolmaker, often your engineer by video |
| T2 | Re-sampling after corrections agreed at T1 | Revised samples, updated report, list of what changed | Same as T1 |
| T3 / pilot run | Short production run to prove stability | Run chart, cycle and weight data, capability study | Toolmaker plus your quality function |
| PP sample / PPAP | Production-part approval, common in automotive | Full submission package, signed off by your customer | Formal, documented, often audited |
| "Final sample" | Not a defined stage — means whatever the writer wants | Ask for the definition in writing | — |
The one that causes real arguments is T0. Some shops call the first closing of the tool T0 and the first customer-facing sample T1; others call the first sample T0 and the correction round T1. If your purchase order says "includes T1 and T2" without a definition attached, you may be paying for two internal attempts and receiving one sample set.
Put this in the PO: the name of each stage, what the supplier delivers at each one, how many are included, and what triggers an additional one. "T1: sample set of 20 parts, trial report, dimensional report on the ten critical dimensions, delivered within X days of mold completion. Correction trial after T1 feedback included." A sentence like that removes the ambiguity completely.
The four things a trial has to prove
Any trial report should be readable as evidence for or against these four claims. If a report makes none of them, it is a photo album with a header.
| What it has to prove | How you verify it | What it looks like when it fails |
|---|---|---|
| The part meets the drawing | Dimensional report on every critical dimension, measured on stable parts | Report lists only "typical" values, or the parts were measured straight off the machine |
| A process window exists | Parameter study at several melt temperatures and injection speeds | Only one recipe produces an acceptable part, and nobody wrote the others down |
| The tool runs continuously | 300–500 consecutive shots with cycle and weight logged | Ten good parts, then a drift nobody can explain |
| A production baseline is transferable | Signed parameter card you can hand to any machine operator | Parameters live in one technician’s head or in a photo of the screen |
The second one is the most commonly skipped, and the most expensive to skip. At T1, a competent toolmaker can usually make a decent part by finding a working set of parameters. That is not the same as the tool being manufacturable. A tool that only produces good parts inside a narrow band will fail as soon as the machine drifts, the resin batch changes, or the shop floor temperature moves ten degrees between seasons — and all three of those happen in normal production.
A supplier who cannot show you a window is telling you something about how they work, not just about this tool. That signal, and the others worth reading at this stage, are collected in how to vet a plastic injection molding supplier.
Measure, do not just look
A trial sample that is visually acceptable can still be unusable, and the reasons are almost always invisible to the eye: a dimension drifting toward the limit, a cavity running slightly heavier than its neighbours, a flatness error that only shows on an assembly fixture.
The list below is what to ask for. It is deliberately short. Measuring everything on a first trial produces a page of numbers that hides the two or three that decide whether the project proceeds.
| Item | How to measure it | Acceptance thinking |
|---|---|---|
| Critical dimensions | CMM, height gauge, pin and thread gauges | Only dimensions marked critical on the drawing. Measuring everything creates noise and hides the one number that matters |
| Part weight | Balance to 0.01 g, ten consecutive parts | The cheapest proxy for cavity-to-cavity balance and for process drift. Spread across a multi-cavity tool should stay inside about 0.5% |
| Flatness and warpage | Surface plate plus feeler gauge, or a fixture | Warpage often appears only after the part cools fully, so measure again the next day |
| Gate remnant | Visual, plus height gauge on the gate area | Must sit inside the drawing allowance. A raised gate on a visible face is a tooling defect, not a trimming problem |
| Flash | Feeler gauge at the parting line | "Light flash" is not a specification. Write a number into the acceptance criteria |
| Color and gloss | Spectrophotometer for ΔE, gloss meter at a fixed angle | Only meaningful if you supplied the color standard and the resin batch is recorded |
| Surface texture | Visual under oblique light, texture plaque comparison | Compare against the agreed plaque, not against a photo on a phone |
| Assembly | Fit the actual mating part or a gauge | A part that meets every dimension can still fail to assemble. Do this before you sign |
| Function | Whatever the part is supposed to do — snap, seal, hinge, clip | Run it at least 50 cycles if it has a living hinge or a snap fit |
Two of these are worth insisting on even when the supplier calls them unnecessary. The first is part weight per cavity: it takes ten minutes on a 0.01 g balance, and it exposes an unbalanced runner, a partially blocked hot runner drop, or a cooling circuit that is not working on one side of the tool — problems that a dimensional report on a single cavity will never reveal. On a hot runner tool there are more places for that imbalance to come from, as the runner comparison sets out.
The second is assembly. A mounting boss can be perfectly inside its tolerance and still interfere with the mating part, because the tolerance stack between two components is not something a single part’s dimension report can see. If you have the mating part or a gauge, bring it, or ask for a fixture to be built.
When to measure: the 24 and 48 hour rule
Plastic parts keep moving after they leave the mold. A part measured ten minutes after ejection and the same part measured two days later are not the same part, and the difference can be large enough to take a dimension from comfortably inside tolerance to outside it.
| Resin family | Settles in about | What the number does over time |
|---|---|---|
| PS, ABS, PC, PC/ABS | 24 hours | Shrinkage after ejection is small; most of it happens in the first day |
| PP, PE, HDPE | 24–48 hours | Semi-crystalline and slow to settle. Post-mold shrinkage continues measurably on day two |
| POM | 48 hours | High and predictable shrinkage, and it keeps moving for two days in thick sections |
| PA6, PA66 (unfilled) | Depends on conditioning | Moisture absorption makes the part grow after molding. Always state whether the measurement is dry as-molded or conditioned |
| PA with glass fiber | 48 hours, plus conditioning | Dimensional change from moisture is smaller but orientation effects are larger |
| PBT, PET | 24–48 hours | Some post-crystallization in the first days; thin walls settle faster |
| PMMA, transparent PC | 24 hours | Dimensionally calm, but internal stress can show up later as crazing |
The practical rule that resolves this:
- Record the immediate measurement. It tells you what the customer will see if they assemble parts straight off the machine, which some do.
- Judge acceptance on the 24 to 48 hour measurement, depending on the resin. This is the number that represents the part as it will exist in a warehouse.
- State the condition in the report. "Measured 48 h after molding, dry as-molded, 23 °C" is a complete statement. "Measured" alone is not.
Polyamides are the trap here. Unfilled PA6 and PA66 absorb moisture from the air and grow after molding — sometimes by a few tenths of a percent, which on a hundred millimeter dimension is the whole tolerance. A supplier who measures dry parts and a customer who measures conditioned parts will argue about it for weeks, and both will be right about their own measurement. Decide up front which condition is the acceptance condition, and write it into the drawing notes.
Why this matters commercially: if the tool is signed off on an immediate measurement and the part keeps shrinking overnight, the correction round becomes your problem to argue about. If the acceptance condition was agreed in writing, the tool is simply not to specification and the correction is the supplier’s cost.
Have a trial report you are not sure how to read?
Send us the report and the drawing. We will tell you which dimensions carry risk, whether the process window is real, and what we would ask the toolmaker before signing. No obligation, and we will say so plainly if the tool looks fine.
The process window test most buyers skip
Ask a supplier "can this tool run in production?" and most will answer by pointing at a good part. The better question is: how far can the parameters move before the part stops being good?
That distance is the process window, and it is measurable with equipment the supplier already has. The method is straightforward: pick three process variables — typically melt temperature, injection speed and hold pressure — set each at a low, nominal and high level, and mold parts across that grid. Then inspect every combination against the acceptance criteria and mark each cell pass or fail.
The result looks like a map. A robust tool has a wide block of passing cells around the nominal setting. A fragile tool has one passing cell, or a diagonal line of them, and everything else failing on short shots at one end and flash at the other.
Why a narrow window is worse than it sounds
Production does not hold parameters still. Screw and barrel wear change the actual melt cushion over months. Resin batch to resin batch variation changes viscosity. A molding shop in an unheated building sees its mold temperature move with the season. A tool with a wide window absorbs all of that. A tool with a narrow window produces a drift of rejects that nobody can explain, and the usual response is to slow the cycle down to make it safe — which quietly destroys the piece price the project was quoted on. The cycle itself is covered stage by stage in the plastic injection molding process.
For automotive work this is formalised as a capability study, with a requirement typically around Cpk 1.33 on critical dimensions. You do not need the full statistical apparatus for a houseware part, but you should still ask for the pass-fail grid. Fifteen extra minutes of molding at T1 buys visibility into the next three years of production.
Reading the trial report line by line
A trial report is a record, not a summary. The difference shows in what is missing. Below is what a complete report contains, and — more usefully — what each omission hides.
| # | What the report should contain | Why the omission matters |
|---|---|---|
| 1 | Tool number, cavity number, part revision | Without these the report cannot be tied to anything later |
| 2 | Resin grade, batch number, drying conditions | Undried resin explains half of all cosmetic complaints at T1 |
| 3 | Machine make, tonnage, screw diameter | You cannot reproduce the result on a different machine without knowing the starting point |
| 4 | Full parameter set: melt temperature, mold temperature, injection speed, hold pressure and time, cooling time, cycle | A report with only temperatures and cycle time is a summary, not a record |
| 5 | Mold temperature controller set point and actual reading | The gap between the two is where warpage problems hide |
| 6 | Date, ambient temperature and humidity | Seasonal drift is real, and it explains why a winter trial does not repeat in summer |
| 7 | Number of parts sampled and when they were taken | Parts from shot 5 and parts from shot 300 are not the same evidence |
| 8 | Dimensional results with the measurement time stated | 1 hour and 48 hours give different answers — see the next section |
| 9 | Part weight per cavity, per shot | The fastest way to spot an unbalanced tool |
| 10 | Cosmetic result, with photos at defined angles | "Acceptable" is not a finding |
| 11 | Open issue list with proposed countermeasures and owner | This is what the next round of work is built from |
| 12 | Anything that was changed on the tool since the last trial | Otherwise you cannot tell what fixed the problem, and it may come back |
The single most common gap is item 4. Many reports list melt temperature, mold temperature and cycle time, and stop there, because those are the three numbers on the machine display that everyone recognizes. The parameters that actually explain a result — injection speed profile, hold pressure and hold time, screw position at transfer, back pressure — are the ones usually left out, and they are exactly the ones you need to reproduce the trial on a different machine.
The dimensional report: which numbers matter
Two things decide whether a dimensional report means anything: which dimensions were chosen, and how much of the tolerance they use.
Chose critical dimensions, not all dimensions
The drawing should already mark them: dimensions that control fit, function, or the customer’s assembly. If the drawing does not, agree the list before the trial, not after. A report with sixty dimensions and no priority forces you to scan for the outlier, which is how real problems get missed. If the drawing is still being finalised, tolerance decisions are better made during DFM review than discovered at sampling.
Tolerance usage is the number to watch
This is the point most buyers never check. A dimension is not simply "in tolerance" or "out". If a 20.00 ±0.10 mm dimension comes back at 20.09, it passes — and the tool is in trouble, because production variation will push it past 20.10 within weeks.
A sensible target at T1 is that critical dimensions sit within the middle half of the tolerance band, roughly 50–60% utilisation. Anything consistently above 80% is a warning that should be resolved before sign-off, even though nothing is technically out of specification.
Direction matters, because steel only comes off
A mold can be adjusted in one direction easily: material can be removed. A cavity that is slightly too small can be opened up by milling or EDM. A cavity that is too large has to be welded and re-machined, which damages the steel locally and is something every toolmaker tries to avoid.
So at T1, external dimensions should be running toward the upper limit and internal dimensions toward the lower limit — leaving room to correct in the direction that is cheap. A toolmaker who delivers every dimension dead center has either been lucky or has not thought about the correction path. Ask which dimensions are expected to move when the tool is corrected, and in which direction.
What cosmetic defects tell you about the tool
At a trial, a cosmetic defect is not just something to be polished out or sprayed over. Each type of defect points at a specific part of the tool, and the useful question is whether the root cause is in the steel or in the settings. If it is in the steel, no amount of process tuning will fix it permanently, and accepting the sample means inheriting the problem.
| What you see | What it usually points to in the tool | Is it the tool’s fault? |
|---|---|---|
| Weld line on a visible face | Gate position or gate count, or an unbalanced runner that splits flow in the wrong place | Yes. This is a design decision, not a setting |
| Raised or stringing gate remnant | Gate type or gate size chosen wrongly for the resin; no gate insert to re-cut | Usually yes |
| Whitened or proud ejector marks | Ejector area too small for the ejection force, or draft too low, so the part is pushed off rather than slid off | Yes |
| Flash on one side of the parting line only | Parting surface fit, or insufficient plate stiffness under clamp load | Yes |
| Burn marks or dark streaks | Missing or blocked venting at the last point to fill | Yes — the tool has no path for the air |
| Sink marks on ribs or bosses | Wall or rib section too heavy in the part design, sometimes paired with short hold | Partly. Correct the part geometry |
| Flow marks or serpentine lines | Gate too small for the fill speed required, or speed too low | Often a gate sizing problem, so yes |
| Cracks or stress whitening at corners | Sharp internal corners, no fillet radius, or off-center ejection | Yes |
The distinction matters for how you respond. A sink mark on a heavy rib is a part-design issue and may be acceptable if the customer agrees. A weld line in a visible location is a gate-layout issue — the tool was designed with the flow splitting somewhere you can see, and only moving or adding a gate will fix it. When a supplier responds to a weld line complaint with "we will adjust the injection speed", the honest translation is that the fix will be marginal and temporary. Gate and runner decisions are made at the mold design stage, which is why they are expensive to revisit later.
Run stability before you sign
Ten good parts prove that the tool can make a good part. Three hundred consecutive good parts prove that it can make them repeatedly, which is the only thing production cares about.
The check is cheap to run and hard to fake: let the tool run for 300 to 500 shots without intervention, and log the cycle time and the part weight every fifty shots.
What you are looking for:
- Cycle time staying within about 5% of its own average. A creeping cycle usually means cooling is degrading, a vent is clogging, or a slide is starting to bind.
- Part weight staying within about 0.5%, with no one-way trend. A steady upward drift suggests a check ring or non-return valve issue; a downward drift often means the melt is becoming inconsistent.
- No step change when the mold reaches thermal equilibrium. Many tools behave differently cold than hot, and the first twenty shots are not evidence of anything.
- No increase in flash over the run, which would point at plate deflection or a hot spot building up.
If the tool is destined for a machine other than the one used at the trial, ask for a short run on a comparable machine, or at minimum a calculation showing the required tonnage and shot capacity with margin. A tool that only runs on the biggest press in the building is a tool with a production constraint built in.
Five hard stops: when to reject
Most trial problems are negotiable. These five are not, because each one will cost more to fix after the tool ships than before.
- Critical dimensions out of tolerance with no correction room. If the steel has already been cut to its limit in the wrong direction, the tool needs welding or a new insert. Do not accept a promise to "manage it in production".
- Only one parameter set produces an acceptable part. No window means no production robustness. This is a tooling or part-design problem that will reappear as a reject rate.
- Cosmetic defects whose root cause is in the tool, answered with process changes. Gate position, venting and draft are not settings. If the answer is "we will adjust the machine", the defect is still there.
- Steel grade or hardness not matching the purchase order. Check the certificate against the PO before the tool leaves, not after a surface starts wearing oddly two years later.
- Cycle or weight drift beyond the limits during the stability run. A tool that will not repeat is not finished, regardless of how good an individual part looks.
A word on the correction round. A T2 is normal and is not a failure. Many tools need one. What is not normal is a T2 that arrives with no list of what was changed. Without that list you cannot confirm the fix addressed the cause rather than the symptom, and you cannot tell whether the process window has actually widened or the problem merely moved.
The handover package
The trial is also the moment to collect documentation, because it is far easier to obtain while the tool is still in the shop that built it. Ask for the package at the same time as the samples, not after the final payment.
| # | Item | What you are protecting yourself against |
|---|---|---|
| 1 | Assembly drawing (2D) and 3D data (STEP) | Being unable to quote a repair, or a copy tool, to anyone else later |
| 2 | Steel certificate and hardness test for cavity and core | A quiet substitution — see the mold steel comparison |
| 3 | Cooling circuit diagram with port sizes and required flow | Nobody knowing which circuit cools what, so the tool runs hot in one area for years |
| 4 | Hot runner zoning diagram, thermocouple type, controller wiring | Buying a controller that cannot talk to the manifold |
| 5 | Ejection and return mechanism description, spring specification | Ordering springs by eye and getting the wrong rate |
| 6 | Spare parts list: ejector pins, springs, seals, hot runner tips, wear plates | A two-week stop for a part that costs less than the courier fee |
| 7 | Machining and heat treatment records | Having no baseline for hardness when someone later questions the tool’s life |
| 8 | Trial report and a retained sample set | No physical reference for what "good" looked like on day one |
| 9 | Production parameter card | Every future setup becoming a rediscovery exercise |
| 10 | Lifting drawing: center of gravity, lifting points, total weight | A crane lift done by guesswork, which is how molds get dropped |
| 11 | Maintenance manual: lubrication points, intervals, wear items | The tool degrading quietly until it flashes everywhere |
| 12 | Machine compatibility table: minimum tonnage, mold height, opening stroke | Finding out on arrival that it will not fit the press you allocated |
Twelve items sounds like a lot until the first time you need one of them. The two that get skipped most often are the cooling circuit diagram and the spare parts list, and both are the ones that hurt: without the circuit diagram, nobody at your plant knows which of the eight water lines cools the core insert, so a warpage problem becomes a trial-and-error exercise every time it appears.
This list also matters if you ever plan to move the tool, sell it, or have a second source quote a repair. A tool without documentation is worth considerably less than one with it, and buyers of used molds know that.
Frequently asked questions
How many trials should a new mold need?
One is possible for a simple part in a familiar resin, but two is the normal case once cosmetic or dimensional corrections are involved. What matters is not the count but the content: each round should close a defined list of issues, and each round should end with a report showing what changed. A tool that reaches T3 with the same open issues it had at T1 is not making progress, and adding more trials will not fix that.
Should I attend the trial in person?
If the part is cosmetic, high value, or has a critical fit, yes — the value is being able to ask "why does it do that?" while the tool is open. If that is not practical, ask for a live video walk-through during sampling rather than a photo set afterwards. Photos are chosen by the person taking them; a live call is not. At minimum, insist that the parameter screen is visible in the video.
What if the parts look good but the report is missing?
Treat it as incomplete, because that is what it is. The report is the only record that lets you tell, months later, whether a new problem is a change or was always there. It is also what you would need if a dispute arises. Asking for a proper report costs nothing at trial stage; reconstructing one after the tool has shipped is usually impossible.
Is a Cpk requirement necessary on a non-automotive part?
Not the full formal study, in most cases. But the underlying question — can this dimension repeat, and how close to the limit does it sit? — is worth answering for any part that will run in volume. A simple substitute is to ask for ten consecutive parts measured on the critical dimensions, with the spread and the position within tolerance stated. If the spread is wide or the position is near the edge, the same concern applies without the statistics.
Who pays for the correction round?
If the correction is for something the tool should have delivered — a dimension out of tolerance, a gate in the wrong place, missing venting — it is the toolmaker’s cost, and that should be stated in the purchase order. If it is a design change you requested after seeing the samples, it is yours. The grey area is cosmetic expectations that were never written down, which is why agreeing the surface standard and the acceptance plaque before the tool is built is worth the effort.
What is the difference between a mold trial and a pilot run?
A trial validates the tool: does it produce a part that meets the drawing, and is there a workable process window? A pilot run validates the system around it: can the line hold that result over a shift, with the real operators, the real packaging, and the real inspection method? Small tools often go straight from trial to production. Anything with volume, multiple cavities or tight tolerances benefits from a short pilot before the full commitment.
