Start with what you can point to.
“The emulsion broke” is a conclusion.
“After 48 hours at 25°C, a clear lower layer measured about 3 mm while the upper phase remained opaque” is an observation.
The second statement gives you something to compare, photograph, measure, and test. It also prevents one early explanation from taking over the investigation.
Describe the defect before you diagnose it.A precise observation can survive a wrong hypothesis. A vague diagnosis cannot.
Rebuild the batch record.
Troubleshooting is limited by the record behind the batch. Capture enough information to reproduce and compare the work.
Materials
Trade name, INCI, supplier, grade, lot, percentage, concentration, storage, and condition before use.
Formula and batch
Formula version, batch size, phases, target and actual pH, final mass or yield, and target performance.
Process
Temperature history, addition order, equipment, speed, time, hydration, cooling, transfers, and deviations.
Result
Appearance, viscosity, pH over time, storage, failure timing, photographs, measurements, and package behavior.
An unspecified detail should remain unspecified. Do not fill a gap with the procedure you usually follow.
Compare before you explain.
A successful control batch is one of the most useful troubleshooting tools. Compare the failed batch with the control and mark each variable as the same, different, unknown, or measured differently.
Look for changes in material lot or grade, percentage, batch scale, addition order, temperature, shear, hydration, cooling, pH adjustment, storage, package, or measurement method. Unknown differences belong on the list too.
Use one repeatable sequence.
- ObserveState the defect in measurable terms. Include timing, location, conditions, and photographs where useful.
- ComparePlace the failed batch beside a successful control or the closest available reference. Mark confirmed changes and missing information.
- HypothesizeRank a small number of competing explanations by how well they fit the observations and known changes.
- TestRun the smallest controlled experiment that can distinguish the leading explanations.
- DecideAccept a change only when it corrects the defect without creating another unacceptable problem.
Possible investigation categories include formulation failure, process failure, measurement artifact, packaging interaction, and microbiological failure. More specific possibilities may include separation, creaming, coalescence, viscosity loss, incomplete hydration, crystallization, aeration, precipitation, oxidation, discoloration, pH drift, or polymer and electrolyte incompatibility. These are leads, not diagnoses.
Why not change several things at once?
A focused one-variable test can answer one immediate question, but it cannot expose every interaction. Hold the control recipe and measurement method steady when that isolates the leading explanation. If several variables may interact, move to a designed experiment instead of treating a series of one-variable tests as complete proof.
Worked example.
Observe
After 48 hours at 25°C, the sample has a clear lower layer measuring about 3 mm. The control remains uniform. The failed sample also has lower measured viscosity, while pH is similar.
Compare
The nominal formula is unchanged. A different supplier grade was used for the emulsifying material, and cooldown agitation ended ten minutes earlier.
Hypothesize
The supplied grade changed the structure, the shorter cooldown process did not build the intended structure, or the viscosity comparison is distorted by a measurement difference.
Test
Confirm the measurement method. Reproduce the control process with the original material, then test the alternate grade under the same process conditions.
Decide
If one controlled change removes the layer and restores the expected measurements without a new defect, it supports the next decision. It does not establish shelf life, preservation, safety, or release readiness.
Professional note
A product can remain physically uniform and still have inadequate antimicrobial protection. It can also pass a preservation assessment and later show a physical, chemical, or packaging problem. ISO/TR 18811 addresses product-specific stability planning. ISO 11930 addresses antimicrobial protection separately. Suspected microbiological failure needs its own qualified investigation.
Quick recap.
- Preserve the actual formula, material, process, measurement, and storage record.
- Describe the observation before naming a mechanism.
- Compare the failed batch with a known-good control.
- Rank competing explanations and run a focused controlled test.
- Keep physical stability, preservation, packaging, safety, and regulatory questions separate.
Sources.
- ISO/TR 18811:2018: Cosmetic stability testing guidance
- ISO 11930:2019: Evaluation of antimicrobial protection
- FDA: Cosmetic GMP guidelines and inspection checklist
- FDA: One-year physical stability study of in-house sunscreen formulations
- NIST: Practical steps for designing experiments
- NIST: Limits of one-factor-at-a-time experiments