How do you tell whether a failed plant trial was valid at all?

This check is for plant trials of industrial chemicals, such as cleaners, coolants, lubricants and water treatment products, judged against criteria fixed before the run. Drug, food and clinical trials follow their own regulations. The FDA guidance cited here governs drug laboratories and is used as a model of reasoning, not as a rule for chemical plants.

What do you need before you judge a failed plant trial?

Before you judge a failed plant trial, you need five records: the trial plan with its success and stop criteria, the sample record, the run log, the control or baseline data and every raw reading. The plan matters most. The NIST/SEMATECH e-Handbook defines an experimental design as "the laying out of a detailed experimental plan in advance of doing the experiment" (NIST/SEMATECH e-Handbook, section 5.1.1, accessed 2026). Criteria written after the result is known cannot decide anything, because they can be bent to fit it.

The sample record names the product, the lot or batch, the delivery date, the certificate of analysis and how the container was stored. The run log gives the dose or concentration, temperature, line speed, line and shift with times. The baseline is data from the same line, measured the same way, without the trial product. Raw readings mean each measurement, not a summary average.

Gather these before anyone argues about the cause. FDA guidance for drug laboratories asks that an assessment start with "no preconceived assumptions as to the cause of the OOS result" (FDA, 2022), and the same discipline protects a plant trial review. Keep retained samples of the product and of the process fluid, stored and handled as the product's safety data sheet (SDS) directs, so that a doubt about the sample can still be tested.

How do you check a failed plant trial step by step?

A failed plant trial is checked in five steps, and the result is read only after the first four pass. Each step compares one record with the plan and writes down the outcome, so the review can be repeated by someone else.

1. Confirm the sample. Match the lot number on the container label, the certificate of analysis and the trial record. Check the storage conditions and age against the technical data sheet. Write the lot on the result sheet.

2. Confirm the dose and run conditions. Compare the logged dose, concentration, temperature and line speed with the plan window, and mark every period outside it with its start, end and size.

3. Confirm the control or baseline. It must come from the same line, the same measurement method and a comparable period. A baseline from another line or season does not count.

4. Confirm data completeness. Check that the planned run length was reached, every planned reading exists and the gauges were checked. The NIST/SEMATECH checklist puts "Check performance of gauges/measurement devices first" at the top of its list (NIST/SEMATECH e-Handbook, section 5.1.3, accessed 2026).

5. Read the result. Compare it with the criteria fixed in the plan, without changing them.

If step 1, 2, 3 or 4 fails, stop. The trial is undecidable, and the remaining sections explain what to record and how to rerun it.

What makes a plant trial invalid rather than failed?

A plant trial is invalid when a documented event other than the product could have caused the result, or a design gap leaves the product's effect unreadable; it has failed only when a valid trial missed a criterion fixed in the plan. The difference rests on evidence, not on how unwelcome the result is. FDA guidance states that invalidating a result "may be done only upon the observation and documentation of a test event that can reasonably be determined to have caused the OOS result" (FDA, 2022). Where an investigation finds no cause and does not confirm the result, the guidance says the result "should be given full consideration" (FDA, 2022). For a plant trial, this suggests a simple rule: no documented event or design gap, such as a missing control or a short run, no invalidation, and the failure stands.

The table below decides whether a failed result can be read as a verdict on the product or must go to a new trial.

The key row is the control. Without a baseline from the same line and method, even a clean run cannot separate the product's effect from run-to-run variation, so the trial goes to a rerun with a control.

How do you check that the plant trial used the right product sample?

You check the plant trial sample by tracing one container from delivery to the line and confirming that its lot, condition and working concentration match the trial record. Start with the paper trail: the lot number printed on the drum or tote, the lot on the certificate of analysis and the lot written in the trial record must be the same. A mismatch is a documented event and enough to stop the review.

Next, check condition. Compare the delivery and opening dates and the storage temperature with the shelf life and storage limits in the technical data sheet. Then test the retained sample from the same container for one or two properties the certificate reports, such as concentration, density or pH, using the method the certificate names. A result outside the certificate value by more than the method's stated reproducibility (the between-laboratory limit) points to the sample, not the product design.

Finally, check what actually reached the process. Many products are diluted on site, so the working solution is the real sample. A titration or refractometer reading of the bath or sump taken during the trial, with the protective equipment named in section 8 of the product's SDS, shows whether the planned concentration was ever reached. FDA guidance describes the same logic for laboratories: an investigation may find that "the original sample was prepared improperly and was therefore not representative of the batch quality" (FDA, 2022). Record each comparison and its result on the trial file.

How do you check that plant trial conditions matched the trial plan?

You check plant trial conditions by comparing the run log with the plan window for each controlled factor, period by period, and listing every deviation with its time, size and duration. Typical factors are dose or concentration, temperature, line speed, load, water quality and the upstream process. A deviation matters when it touches a factor known to move the result.

Knowing which factors move the result is the hard part. ASTM E1169-21(2026), a standard practice for ruggedness tests, exists "to identify those factors that strongly influence the measurements" and "to estimate how closely those factors need to be controlled" (ASTM E1169-21(2026)). The practice addresses test methods, but the question is the same on a production line. If the trial plan never ranked the factors, treat any deviation in temperature, concentration, contamination or the upstream process as possibly decisive.

Look for unplanned changes as well as drift. The NIST/SEMATECH checklist warns: "Watch out for process drifts and shifts during the run" and "Avoid unplanned changes (e.g., swap operators at halfway point)" (NIST/SEMATECH e-Handbook, section 5.1.3, accessed 2026). Ask maintenance and production about repairs, other chemical changes, raw material lot changes and shift changes during the trial. A deviation that also occurred in the control period weakens the comparison less than one that hit only the trial period.

What do you do with a plant trial that cannot be judged?

A plant trial that cannot be judged is recorded as undecidable, with the check that failed, the evidence and the change the rerun must make, and it is not reported as a product failure. Writing "undecidable" protects both sides: the product is not rejected on a broken test, and it is not approved on one either.

Two shortcuts make it worse. Averaging the original result with rerun results hides the problem; FDA guidance says such averaging "is not appropriate because it hides variability among the individual results" (FDA, 2022). Rerunning until one run passes is the pattern FDA calls "testing into compliance" (FDA, 2022). Report every run on its own, and decide in advance how many reruns the plan allows.

Design the rerun to close the gap and nothing else: verify the lot, add the control, log the conditions and run to full length, with the original criteria unchanged. The NIST/SEMATECH e-Handbook advises that "Planning to do a sequence of small experiments is often better than relying on one big experiment to give you all the answers" (NIST/SEMATECH e-Handbook, section 5.1.3, accessed 2026). Lubinpla, an applied AI company for specialty chemical technical reasoning, works such a case from your own trial documents and site history, judges the trial's validity before any single result and leaves an invalid trial as undecidable for a new trial, in one cited document for your review.