Stability Testing: What to Ask For and What You Get

Sourcing Guide · 2026-10-21

The question that outlives the launch

Most sourcing conversations are about how a product looks and performs on the day the sample is approved. Stability is about what it looks and performs like in month twelve, in a warehouse that is hotter than the label assumes, in a pack that has been sitting on its side.

It matters commercially because the failure arrives late. By the time a product separates, discolours or starts to smell different, you have printed the cartons, shipped the stock, and made a shelf-life claim to a retailer. Every one of those is expensive to unwind, and none of them is fixable by a phone call to the factory.

So stability is worth asking about early — not as a compliance box, but because the answer tells you how the factory thinks about risk.

What stability testing is, and what it is not

Stability testing is a structured observation programme. The same batch is placed under defined conditions, defined properties are measured at defined intervals, and what changed is written down. That is the whole idea: a record of change over time, under conditions you can name.

It is not one test, and it does not produce a certificate. There is no universal "passed stability" stamp that travels with a formula from factory to factory. A report describes a specific batch, in a specific pack, under a specific protocol — change the fragrance dosage or the bottle and the old data no longer describes what you are selling.

It is also not a safety or efficacy study. Stability data says whether the product holds together physically, chemically and microbiologically, and whether the pack survives contact with it. Preservative efficacy is a separate test with its own protocol, and claims about what the product does for skin or hair are supported by different evidence again.

The parts of a normal programme

Programmes differ by product and by market, but the components are recognisable. When a factory describes its stability work, these are the pieces to listen for:

• Elevated temperature, usually around 40 °C for a set period, with samples pulled at intervals
• Real-time storage at the conditions the product is expected to live in — often quoted as 25 °C / 60 % RH, or 30 °C / 65 % RH for hotter climates
• Freeze–thaw cycling, for products that will cross a cold supply chain in transit
• Light exposure, for anything sold in a clear or translucent pack
• Pack compatibility: the product in the actual bottle, tube or jar, not only in glass
• Tracking of the properties that matter for the format — appearance, pH, viscosity, colour, odour, separation, and for emulsions the state of the dispersed phase
• A preservative efficacy or challenge test, run on the formula and re-checked if the pack or the preservation strategy changes

The pack point is the one buyers most often miss. A formula can be perfectly stable in a glass beaker and still fail in an airless bottle: the pump can seize, the lining can be attacked, the label can lift, and water can leave through the wall over a year. Testing in the commercial pack is what turns a laboratory result into a shelf-life decision.

Accelerated data predicts; real-time data confirms

Accelerated testing exists because nobody wants to wait three years to find out. Raising the temperature speeds up the reactions that degrade a formula, and the pattern of change is used to estimate what will happen at normal conditions.

That estimate is useful and it is not proof. The relationship between temperature and degradation is not the same for every reaction, and some failures — a slow pack interaction, a gradual emulsion drift, a preservative being consumed — do not accelerate on the same curve. Accelerated data is best read as a screen: it finds the formulas that will not survive, quickly and cheaply. Real-time data is what eventually supports the shelf life you print.

In practice many products launch on accelerated results with a real-time study still running. Whether that is acceptable for your product and your market is not a decision the factory can make for you — it depends on the destination rules and on what your buyer will accept. Ask both.

One more thing worth saying plainly: a formula that fails at the accelerated stage has failed cheaply. That is the moment the formulator can still change the emulsifier, the preservative or the pH, before tooling and artwork are committed.

Ask for the protocol before you ask for the result

A stability claim without a protocol behind it is an opinion. Before you look at any conclusion, ask how the study was set up. These are the questions that surface whether there is a programme or a habit:

• Which conditions, at which temperature and humidity, and for how long?
• How many timepoints, at which intervals, and what is measured at each one?
• Was the sample in the final pack, or in glass?
• How many batches were tested — one, or more than one?
• Who performs the work: an in-house laboratory or an external one?
• What is the acceptance criterion for each property, and was it written down in advance?

The last question is the one that separates a test from a judgement. "No significant change" means nothing unless someone defined significance before the first timepoint. A pH that moves by 0.3 may be irrelevant in one formula and a warning in another; a colour shift that is a defect in a clear gel may be acceptable in an opaque cream. If the criteria were set after the data came in, the study was designed to pass.

Batch count matters for a different reason. One batch tells you about that batch. More than one tells you something about the process — whether the plant can make the same product twice, which is the thing you are actually buying at volume.

What you should receive at the end

A stability file is a small stack of documents, and you should ask for the stack rather than the summary sentence. What a usable package contains:

• The protocol: conditions, duration, timepoints, properties measured, acceptance criteria
• The batch reference, tied to a production record you can trace
• The raw results per timepoint, not only the conclusion
• Photographs of the samples at each timepoint, taken under the same conditions
• Any deviation — a missed timepoint, a sample stored wrongly, an instrument recalibrated
• The conclusion: recommended shelf life, storage conditions, and any restriction on the pack

A single line reading "stable for 36 months" with no tables behind it should be treated as a claim, not a result. It may well be true; you simply have no way of knowing, and neither will your importer if they ask.

What failure looks like

Failure is usually gradual and easy to miss if nobody is measuring the right property. The common patterns are worth knowing by name, because they tell you what to look at in the report:

• Separation, creaming or sedimentation in an emulsion
• pH drift, which often signals a preservative losing effectiveness before anything is visible
• Viscosity change — thinning in surfactant systems, thickening or gelling in others
• Colour shift or fading, especially with natural extracts and some fragrance components
• Odour change, which is frequently the first sign of oxidation
• Pack interaction: lining attack, corrosion at the closure, a pump that stiffens, a label that lifts, weight loss through the container wall
• Microbial growth at a late timepoint, which points back to preservation rather than to the emulsion

Not every change is a failure. A natural extract may shift colour slightly and still be saleable if the criterion allowed for it and the label does not promise a fixed shade. The problem is not change; it is change nobody agreed to in advance.

Stability is a formulation decision, not a final inspection

Buyers sometimes treat stability as the last gate before shipping. It is closer to the opposite: it is a consequence of decisions taken at the bench, and the further a project gets before those decisions are examined, the more expensive the correction.

The variables that most often move a stability result:

• The preservative system, and whether it is effective at the pH the formula actually sits at
• pH itself, since both extremes make life harder for emulsifiers and for packaging
• Natural extracts and actives, which bring colour, odour and degradation pathways with them
• Electrolyte level, which can destabilise an emulsion that looked fine in the beaker
• Fragrance dosage, which can both destabilise a base and discolour it over time
• Pack format: an airless bottle, a jar and a tube expose the same formula to very different conditions

This is why the brief should carry any claim you intend to make. If the product has to be free of a class of ingredient, or has to look a particular way, say so before the stability batch is made — not after the sample has been sitting at 40 °C for a month.

How it fits into the calendar

Stability runs on its own clock and does not compress. The real-time study starts when the batch is made, which means it is usually still running when the first production order ships. That is normal, and it is why the launch plan should be built around accelerated results with the real-time study continuing in the background.

Two practical consequences. First, the shelf life on the label should be a decision made with the stability conclusion in hand, not a number chosen to match a competitor — and the storage statement on the pack should match the conditions that were actually tested. Second, if your market requires stability data as part of a registration or notification file, the study must start early enough to be finished before you need the file. That lead time is worth confirming at the quoting stage, when there is still room to move.

A short request list to send the factory

If you want to test how a factory handles this without asking for a full report, send these seven lines. The quality of the reply tells you as much as the content:

• Please send the stability protocol you would use for this product and pack.
• Confirm the conditions, the duration and the timepoints.
• Confirm whether samples are tested in the final pack or in glass.
• Confirm how many batches are included.
• Send the acceptance criteria for each property measured.
• Confirm who performs the testing, and whether the raw data is available to us.
• Tell us what happens if a timepoint fails — who decides, and what the options are.