Your parts passed inspection, shipped, and went into your product. Weeks or months later, the reports start coming in: a housing cracked at a screw boss, a clip snapped during assembly, a cover warped so it no longer seals. The parts were approved. The mold was approved. And now they are failing in your customers’ hands.
Contain the problem first, then preserve the evidence, then find the cause — in that order. Stop more bad parts from reaching customers, keep failed parts and good parts from the same lot untouched, and only then ask the supplier for a written corrective action with evidence. A failure investigated in the wrong order — cause first, containment later — costs you customers while you are still arguing about whose fault it is.
How to use this guide: the first table is the response sequence — work it top to bottom. The triage table tells you which cause family each symptom points at and what evidence separates them. The evidence-pack checklist is what to send your supplier so the investigation can actually start.
First, contain the problem
Containment is not the same as finding the cause. It is the work that stops the damage from growing while the cause is still unknown. The eight disciplines (8D) method, the corrective-action format used across manufacturing supply chains, makes the same distinction: as the American Society for Quality (ASQ) explains, the 8D report establishes a permanent corrective action by determining the root cause — but its third discipline, D3, is an interim containment action that must protect the customer from receiving more defective product until that permanent action is in place. Containment first, root cause second.
| Step | What to do | What not to do |
|---|---|---|
| 1. Stop the flow | Quarantine the affected lot — at your warehouse, in transit, and at the supplier if more parts are still being made. Tell the supplier in writing, with the lot number and the failure, and ask them to hold the lot. | Do not keep shipping "just the good-looking ones" from a lot that has failures. If you cannot tell good from bad by inspection, the lot is not shippable. |
| 2. Count the exposure | How many parts are affected, where are they, and which customers have them? Separate three groups: parts in your stock, parts in your customers’ hands, and parts still at the supplier. | Do not guess the scope from one report. One cracked housing may be the first of a lot-wide problem or a single damaged part — you cannot know until you count. |
| 3. Decide on the field action | For parts already with customers, decide — with your own team, not the supplier alone — whether the failure needs a replacement, a repair, a warning, or a recall. Safety-critical failures are a different decision from cosmetic ones, and the decision belongs to you. | Do not let the supplier decide the field action for you. They can advise on the cause; the risk to your customers is yours to judge. |
| 4. Put it in writing | One short written notice to the supplier: what failed, when, how many, which lot, what you have contained, and what you need next. Date it. This notice starts the clock on the corrective action. | Do not run the whole investigation by phone or chat. A failure with no written record becomes a disagreement about what was said. |
Two points about containment that buyers often get wrong. First, sorting and rework are containment, not a fix. Re-inspecting a lot and removing the visibly cracked parts protects the next customer, but it does nothing about the parts that will crack next month — the ones that look fine today. Second, containment has a cost, and that cost belongs in the conversation with the supplier from the start. Sorting, rework, replacement parts, freight, and your own inspection time are all part of the failure’s cost. Keep a record of them; whether the supplier pays is a commercial question for later, but a cost with no record is a cost nobody reimburses.
Preserve the evidence before you "fix" anything
The most common mistake in a field failure is also the most human one: someone throws the broken part away, or "fixes" the assembly, or returns the whole lot to the supplier for credit — and with it goes the only evidence of what happened. A failure investigation runs on physical evidence. Without it, every theory is equally plausible and none of them can be proven.
Keep four things, untouched:
- Failed parts, as found. Do not clean them, do not glue them, do not cut them open. The fracture surface — where the crack started, and how it spread — is the single most informative piece of evidence. A crack that started at a sharp corner tells a different story from one that started at a weld line, and both are invisible once the part is cleaned or reassembled.
- Good parts from the same lot. A failed part alone proves something broke. A failed part next to a good part from the same lot, same cavity, same material batch proves what changed. Keep at least a few good parts, identified by lot and cavity if your parts carry cavity marks.
- Parts from a lot that did not fail. If an earlier lot worked and a later lot failed, the comparison between the two lots is where the cause usually hides: a different material batch, a different machine, a tool change, a process change, a new operator. Keep retained samples from earlier lots if you have them — and if you do not, note that gap, because it limits what the investigation can prove.
- The records. The T1 approval report, the material certificate and lot number, the process record for the failing lot, the inspection report, and the assembly instructions the part was used with. If the supplier cannot produce the process record for the failing lot, that itself is a finding: a process with no record cannot be shown to have been in control.
Label everything with the lot number, the date, and the cavity if known. Photograph the failure before anything is moved — the crack, the location on the part, the assembly it came from, and the environment it was used in. A photo taken on the day of the failure is evidence. A photo taken after the part has been handled, cleaned, and shipped back is a picture of a part.
What the symptom tells you — and what it does not
Plastic part failures fall into a small number of cause families, and the symptom narrows the field — but it does not identify the cause on its own. Empire Precision Plastics, a US molder in Rochester, New York, puts it the same way in its failure-diagnosis guide: assuming the part was properly designed, the root cause usually lies in material handling and storage, conditions within the molding cycle, post-molding handling and finishing, or the application and end use — and failures can be classed as mechanical, thermal, chemical, or environmental. The table below follows that structure, and adds the evidence that separates one family from another.
| Symptom | Likely cause families | Evidence that separates them |
|---|---|---|
| Crack at a boss, rib, or screw post | Design (sharp corner, thin section, stress concentration); assembly (over-torque, press fit, screw too long); process (weld line at the failure point, residual stress); material (wrong grade, degraded resin) | Where did the crack start — at a corner, at a weld line, or at the screw? Does it fail at the same place every time? Compare the torque and screw used against the design. Check whether the failure is cavity-specific. |
| Brittle fracture with little deformation | Material (degraded resin, wrong grade, contamination, excessive regrind); process (excessive shear, overheated melt, poor drying); service (low temperature, chemical attack) | Did earlier lots pass the same test? Check the material lot certificate and whether regrind was used and how much. Ask for the drying record and the melt-temperature record for the failing lot. |
| Warpage or dimensional drift weeks after delivery | Process (uneven cooling, packing, gate freeze); material (moisture absorption, continued shrinkage after molding); tool wear (cooling or venting changed); storage (heat, stacking load) | Measure the same parts now and compare with the inspection report at shipment. Did the parts move, or were they always out of tolerance? Check whether the drift is the same across cavities. |
| Discolouration, surface change, or crazing | Material (UV, heat, chemical exposure); process (burn marks, contamination, poor drying); environment (cleaning chemical, weathering, ozone) | When did the change appear — at delivery or after use? What chemicals, cleaners, or temperatures has the part seen? Compare with a part kept in storage. |
| Parts pass inspection, then fail in assembly or use | Design (tolerance stack, wrong datum, weld line in the load path); inspection gap (the failing feature was never measured); end use changed (higher load, new chemical, hotter environment) | Was the failing feature on the drawing, and was it measured? Has the end use changed since the part was approved? Test a retained sample from the approved lot under the same conditions. |
Read the table as a way to ask better questions, not as a diagnosis. A crack at a boss can be a design problem, an assembly problem, or a material problem — and the difference is decided by evidence, not by which explanation is offered first. Three questions from Empire’s Failure Analysis Inquiry are worth asking in every case, because they separate the families fastest: did the part work properly before, when did the problem start, and is the problem cavity-specific? A failure that started suddenly, after months of good parts, points at something that changed — a material lot, a machine, a tool repair, a process setting. A failure that has been present since the first lot points at the design, the material choice, or the approval evidence.
One cause family deserves a note of its own: tool wear. A mold that has run many cycles can produce parts that slowly drift out of specification — flash grows, dimensions move, the surface degrades — and the drift is easy to miss lot by lot. Xometry’s injection mold maintenance checklist makes the point directly: poorly maintained molds may produce parts that fail to meet specifications, and maintenance activities should be logged so faults can be traced to their root causes. If your failure appeared gradually, ask for the maintenance log and the shot counter reading alongside the process record. Our mold maintenance agreement guide explains what maintenance records a buyer should ask for before production starts — and a field failure is the moment those records earn their keep.
What to send the supplier: the evidence pack
A supplier cannot investigate a failure from a description. They need the evidence pack — and the quality of the pack decides the quality of the answer. Send the pack with the written notice, not after a week of emails asking for it piece by piece.
- Failed parts — as found, labelled with lot, date, and cavity. At least three failed parts if you have them; one part is a anecdote, three is a pattern.
- Good parts from the same lot — labelled the same way, so the supplier can compare.
- Photos — the failure as found, the fracture origin close up, the assembly the part came from, and the service environment.
- Lot identification — part number and revision, lot number, production date, cavity number, material grade and lot number, colour batch.
- Service conditions — temperature, load, chemicals and cleaners, UV exposure, assembly torque and method, how long the part was in service before it failed.
- What changed — anything that changed since the last good lot: a new material batch, a new machine, a tool repair, a design revision, a different assembly step, a different end use.
- Your records — the T1 approval report, the inspection report for the failing lot, and the drawing revision the parts were made to.
- What you need back — state it in the notice: containment confirmation, the root-cause investigation, the corrective action, and the evidence that the fix works. Ask for dates, not assurances.
If you cannot supply part of the pack — no retained samples, no material lot number, no T1 report — say so in the notice. A missing record is itself information: it tells both sides what the investigation cannot prove, and it tells you what to fix in your own records before the next order.
What to demand: a written corrective action, not an apology
A supplier who says "we will be more careful" has not corrected anything. What you need is a written corrective action — and the standard format for one is the 8D report. ASQ describes the 8D report as a problem-solving approach that establishes a permanent corrective action and focuses on the problem’s source by determining its root cause, and its eight disciplines run in order: form a team (D1), describe the problem in measurable terms (D2), contain it (D3), find and verify the root cause (D4), choose permanent corrective actions (D5), implement and validate them (D6), prevent recurrence (D7), and close the report (D8). Ask for the report by name, and judge it by three tests:
- Does it name a verified root cause, or a theory? A root cause is verified when the evidence supports it and the other families have been ruled out — not when it sounds plausible. "The resin was wet" is a theory until the drying record or a moisture test supports it. "The weld line is in the load path" is a theory until a part sectioned at the weld line shows it. If the report’s root cause has no evidence attached, send it back.
- Does the fix remove the cause, or add an inspection? Adding an inspection catches the next failure; it does not prevent it. A permanent corrective action changes the design, the material, the process, the tool, or the procedure so the failure cannot recur. Inspection is containment. It is not a corrective action.
- Does it prove the fix worked? The supplier should show measured evidence from the corrected process — parts from the next consecutive lots, measured against the same drawing, with the same failure mode tested. One good sample proves nothing; a run of good parts, measured and recorded, is the beginning of proof. Ask for capability evidence across consecutive lots on the critical dimensions, not a single passing sample.
Set a date for the report in your original notice, and a date for the re-qualification evidence. If the supplier misses the date, that is a finding about the supplier, not just about the failure. And keep the report: it is the record that the failure was investigated, what was found, and what was changed. The next time the same part fails, the first question is whether the same cause came back — and the 8D report is how you answer it.
Who pays, and when to re-qualify or leave
Who pays for a field failure is a commercial question, and the honest answer is that it depends on what your contract, quotation, and quality agreement actually say — and on what the root cause turns out to be. Three positions cover most cases, and you should know which one you are in before the negotiation starts:
- The cause is the supplier’s process, material, or tool. A wrong material batch, an unapproved regrind level, a process run outside the agreed window, a worn tool that was not maintained — these are the supplier’s responsibility, and the supplier should bear the cost of containment, replacement parts, and the corrective action. The evidence pack is what proves it.
- The cause is the design, the specification, or the end use. A sharp corner the drawing required, a load the part was never specified for, a chemical the supplier was never told about — these are the buyer’s responsibility, and the fix is a design change, which our design-change guide covers stage by stage. The supplier can still help, but the cost of the change is a negotiation, not a warranty claim.
- The cause is shared or unclear. A tolerance that was achievable but never measured, a test the part was never put through, a requirement that was discussed but never written down. These are the most expensive failures, because each side can reasonably blame the other — and they are the reason the evidence pack and the written records matter. If the cause cannot be proven, negotiate the cost split on the evidence you do have, and fix the records so the next failure is provable.
Re-qualification is the step buyers most often skip. After the corrective action, the corrected parts need to be approved again — against the same drawing, with the same inspection, and with the failure mode specifically tested. Our T1 approval guide explains how a buyer approves samples before production release, and the same discipline applies after a field failure: approve the corrected parts before volume resumes, and keep the approval on file. A field failure that is "fixed" without re-approval is a field failure waiting to happen again.
And if the supplier cannot produce a root cause, or will not put the corrective action in writing, or the same failure comes back after a fix — that is a supplier decision, not a quality decision. A supplier who cannot investigate a failure cannot prevent the next one. At that point the question is not how to fix the part, but whether to move the tool, qualify a second source, or both.
What a field failure often reveals about approval
It is worth asking, once the immediate problem is contained, whether the failure should have been caught before the parts shipped. Not to assign blame, but because the answer decides what changes in your own process:
- Was the failing feature measured at approval? If the crack started at a boss that was never dimensioned on the drawing, the approval could not have caught it. The fix is a drawing change, not a supplier change.
- Was the failure mode tested? A part that passed dimensional inspection and then cracked under load was never tested under load. Dimensional approval is not functional approval. If the part carries load, the approval should include a functional test.
- Did the approved sample represent production? T1 samples are made under trial conditions. If the production process drifted from the trial — different machine, different settings, different material batch — the approval evidence describes a part you no longer receive. The process record for the failing lot, compared with the trial record, answers this.
- Did anything change after approval? A material substitution, a regrind level, a tool repair, a gate change — any of these can invalidate the approval evidence. One common change is unapproved regrind mixed into virgin resin, which changes impact strength and dimensional behaviour; it should be agreed in writing before it happens, not discovered after a failure.
Illustrative example (not a Cavity Mold project): a buyer received housings that cracked at a screw boss three months after delivery. The crack started at a sharp internal corner, not at a weld line. The material lot had changed between the approved lot and the failing lot, but the corner radius had been the same from the start — and the assembly torque had been increased by the buyer’s own team in the same period. The investigation needed all three facts before it could separate them: the corner was a design weakness, the torque increase was the trigger, and the material change was a coincidence. Any one fact on its own would have pointed at the wrong fix.
Frequently asked questions
Should I return the failed parts to the supplier?
Send some, not all. Keep at least one failed part and one good part from the same lot under your own control. If the supplier holds all the evidence, you cannot check their conclusion — and if the parts are lost in transit, the investigation has nothing to work from.
How long should a corrective action take?
It depends on the failure, and no fixed number would be honest. What you can require is a dated plan: containment confirmed by a date, root cause by a date, corrective action and re-qualification evidence by a date. A supplier who will not commit to dates has not staffed the investigation.
The supplier says the parts met the drawing. Is the case closed?
No. Meeting the drawing proves the parts matched the specification; it does not prove the specification was right. If the part fails in its real use while meeting the drawing, the drawing, the test, or the end-use description is incomplete — and that is a design and specification question, not a closed case.
Can a field failure be the buyer’s fault?
Yes — and the evidence pack is how you find out. A load the part was never specified for, a chemical the supplier was never told about, an assembly torque above the design limit, or storage in heat the part was not rated for are all buyer-side causes. Finding that out is a good outcome: it means the fix is in your control.
What to send, and where to send it
If your molded parts are failing in the field, send the evidence pack — not just a description — and ask for a written corrective action with dates. Through our contact page, send the failed parts and good parts from the same lot (or photographs if the parts are already with your customers), the lot and cavity identification, the material grade and lot number, the service conditions and assembly method, what changed since the last good lot, and your T1 approval report. Tell us what the failure looks like, when it started, and how many parts are affected — and we will tell you what the evidence points at, and what to ask for next.

