How do you run a root cause analysis on a chemical complaint?

This method covers a complaint about a delivered chemical product, such as a coating, adhesive, lubricant or process fluid, where the question is why that lot performed as it did. It does not replace a regulated recall procedure, a process safety incident investigation or the 8D report format your customer may require.

What do you need before a root cause analysis on a chemical complaint?

Before a root cause analysis on a chemical quality complaint, you need the complaint in the customer's own words, the lot number, and the records that can test a cause: the certificate of analysis (CoA), the batch record, a retained sample and the technical data sheet (TDS). ISO 10002:2018 frames complaint handling as a process that includes analysing and evaluating complaints to improve products and services (ISO, 2023), so the analysis starts from the complaint record, not from a meeting. Ask the customer for photos, the application conditions and any test they ran. Check whether a retained sample of the complained lot exists before anyone discards material. US drug rules show why it matters: 21 CFR 211.170 requires a reserve sample representative of each lot in each shipment of each active ingredient (eCFR, 2026). A sealed retained sample of a chemical lot lets you test a hypothesis instead of arguing about it. Collect the shipping date, storage conditions and the container open date. Name who decides the verdicts and who approves the reply.

How do you separate what the customer observed from what was reported secondhand in a chemical complaint?

To separate what the customer observed from what was reported secondhand in a chemical complaint, write every statement on its own line and tag it with who saw it and how they know. Step 1 of the method is this split. A purchasing manager who writes "the coating failed" is reporting secondhand; the line operator who saw tack after 24 hours at 15 °C (illustrative) is the observer. HSE guidance on investigations says to talk to everyone nearby, "especially those who saw what happened" (HSE, 2004). Use three tags: observed by a named person, measured with a stated method, or reported by someone who did not see it. Then ask one follow-up question per secondhand statement, aimed at the person who saw it. Keep interpretations out of the observation column: "the product is off-spec" is a conclusion, while "the film stayed soft after the stated cure time" is an observation. HSE also advises investigators to avoid "making snap judgements on the probable causes" while gathering facts (HSE, 2004). A complaint with only secondhand statements after the follow-up is still workable, but every verdict built on it states that limit.

Why should root cause hypotheses be fixed before you look at the evidence?

Root cause hypotheses should be fixed before you look at the evidence because evidence read first tends to be read in favor of whichever cause the team already suspects. Step 2 of the method is writing the full list of candidate causes before anyone pulls the CoA or retests the retained sample. Nickerson's review in Review of General Psychology defines confirmation bias as "the seeking or interpreting of evidence in ways that are partial to existing beliefs, expectations, or a hypothesis in hand" (Nickerson, 1998). In a chemical complaint, the usual suspect is either the lot or the customer's application, depending on who runs the analysis. Cover both sides and the path between them: raw material or batch deviation, packaging, transport and storage, shelf life, the customer's mixing or dilution, substrate and surface preparation, equipment, and environmental conditions. HSE guidance says the analysis "must consider all possible causes" and advises: "Do not reject a possible cause until you have given it serious consideration" (HSE, 2004). Date the list and save it. Adding a hypothesis later is allowed, but record when and why it was added, so a reviewer can see the list was not rebuilt to fit the answer.

What makes a root cause hypothesis for a chemical complaint disconfirmable?

A root cause hypothesis for a chemical complaint is disconfirmable when you can name, in advance, a result from a record or test that would show it is wrong. Step 3 of the method sets that prediction for every hypothesis on the list. "Something went wrong in production" fails the test, because no batch record or lab result could ever rule it out. "The lot was below the viscosity specification" passes, because a retest of the retained sample can contradict it. The idea comes from Karl Popper: the Stanford Encyclopedia of Philosophy summarizes his view that evidence should count only if it is the positive result of a "risky" prediction "which might conceivably have been false" (Stanford Encyclopedia of Philosophy, 2026). For each hypothesis, write two lines: what you expect to see if it is true, and what result would rule it out. Prefer checks that split hypotheses: a retained sample that performs correctly under the TDS conditions weakens a lot defect and strengthens an application or storage cause. If you cannot write a disconfirming check, narrow the hypothesis until you can, or mark it as one that this analysis cannot test.

How do you record a confirmed, ruled-out or open verdict for each root cause hypothesis?

Record a confirmed, ruled-out or open verdict for each root cause hypothesis by writing the verdict next to the prediction, the check that was run, and one sentence of reason that names the record. Step 4 tests the evidence and step 5 records the verdict. The table below decides which candidate cause of a chemical quality complaint is confirmed, ruled out or left open, using an illustrative complaint about an adhesive lot that did not cure on the customer's line.

The key row in the root cause verdict table is the lot specification row: when the retained sample passes under the TDS conditions, a defect in the lot as released is ruled out for the property tested; packaging, transport and storage stay on the list until their own checks run, and the customer's process moves up. An open verdict is a valid result when the deciding record is missing or cannot be tested. Write which record would settle it, who holds that record and the date you asked for it, so the customer reply can state what would close the hypothesis.

How do you check that a chemical complaint root cause analysis is complete?

A chemical complaint root cause analysis is complete when every hypothesis on the dated list has a verdict and a reason, every confirmed cause explains all of the observed symptoms, and the scope question has an answer. Check four things. First, no hypothesis is missing a verdict; a cause that was dropped without a reason is a gap, not a ruling. Second, the confirmed cause explains every observation tagged in step 1, including the ones that did not fit the first theory. Third, the analysis asked whether other lots or other customers could be affected. In US drug manufacturing, 21 CFR 211.192 requires the investigation to extend to other batches and products that "may have been associated with the specific failure or discrepancy" and requires a written record that includes the conclusions and follow-up (eCFR, 2026). A chemical supplier outside that rule can still use the same scope question. Fourth, a person other than the analyst has read the record and agrees that each reason follows from the cited record. If two causes are confirmed, keep both: a complaint can have a lot cause and an application cause at once.

Common mistakes in chemical complaint root cause analysis

Common mistakes in chemical complaint root cause analysis come from skipping one of the five steps or from treating a test result as more than it shows.

• Starting with the CoA and building hypotheses around what it shows, which lets the first record set the list.

• Accepting the purchasing summary as the observation and never asking the person who saw the failure.

• Writing hypotheses no result could disprove, such as "process variation", so they can never be ruled out.

• Retesting until a result passes and reporting only the passing result.

• Closing a hypothesis as ruled out when the deciding record was missing; that verdict is open.

• Reporting one root cause when the evidence supports two.

• Discarding the retained sample or the customer's remaining material before the verdicts are recorded.

• Following a generic technique such as the 5-Why method or a fishbone diagram without a prediction for each branch. The ZDHC Roadmap to Zero guidance lists fishbone diagrams, 5-Why and additional sampling and testing as options (ZDHC, 2026); each still needs a check that could prove the branch wrong.

When should you bring a specialist into a chemical complaint root cause analysis?

Bring a specialist into a chemical complaint root cause analysis when the deciding check needs a test method your lab does not run, when the complaint involves injury, a regulatory report or a recall decision, or when two hypotheses remain open after the records are exhausted. An analytical lab can run comparative tests on the retained sample and the customer's material, such as infrared spectroscopy for contamination or rheology for cure, under a method you agree before the test. A process safety incident follows its own investigation rules and team, and a drug or food contact product may carry reporting duties under the rules of its jurisdiction, so involve the regulatory owner first. When the deciding record sits with the customer, ask for a joint review with their quality engineer. Lubinpla runs root cause analysis this way: it fixes the candidate causes before it looks up evidence, checks them against your own records such as TDS, SDS, CoA and field history and then against public sources, records a verdict and reason for each hypothesis, and asks for your approval at each step. Bring the dated hypothesis list and verdict table.