In this article
The short answer
Run cosmetic tests in this order: freeze the formula, pack, process and claims; have the safety assessor review it on paper; make one representative batch in the final pack; check its microbial count; start stability, packaging compatibility and the ISO 11930 challenge test on that same batch; let the assessor sign the CPSR on the challenge result and enough stability data; and only then run volunteer tests such as HRIPT and claims tests. Change the formula after testing starts and stability usually has to start again.
Key takeaways
- The law sets no test list, but every result has to describe the product you sell. The Commission guidelines ask for evidence that the stability sample matches the marketed product.
- Freeze four things before booking a lab: the formula with suppliers, the primary pack, the process and the claims list.
- Start with a paper review by the assessor. It removes tests you don't need and stops you testing a formula that has to change.
- Use one representative batch for every test. For a 50 ml cream that often means 30 to 40 filled units plus a glass control set.
- Clear the batch with a microbial count, then start stability, compatibility and the challenge test together.
- Volunteer tests such as HRIPT only confirm safety once the product has been assessed as safe; they are not a way to find out.
- A formula change almost always restarts stability. A pack change usually restarts compatibility and stability in the new pack, but not always the challenge test.
A client of ours once paid for every test twice. She had a water based body lotion ready, so she sent the lab a jar from a kitchen sized batch for a challenge test. It passed. Three weeks later she changed the jar for a pump bottle, swapped the fragrance for a cheaper version from another supplier, and then booked stability. At the eight week reading the pH had dropped from 5.6 to 4.9 and the lotion smelled faintly sour. The challenge test now described a product she was no longer making, and the stability study described one she couldn't sell. Nothing was wrong with any single test. The order was wrong.
This guide is about that order. It doesn't repeat what each test is; we have separate guides to stability testing, the challenge test, microbiological testing and patch testing, and to whether your product needs lab testing at all. If what you want is a calendar, our guides to how long a CPSR takes and the real EU launch timeline cover durations. Here we deal with the sequence: what to freeze first, how many units to make, which tests can run side by side, and exactly what you have to repeat when one of them fails.
Why the order matters at all
Neither the GB Cosmetics Regulation nor the EU Regulation (EC) No 1223/2009 sets a list of tests. What Article 10 and Annex I require is a safety report built on data about the product you actually sell: its stability, its microbiological quality, its packaging, its exposure and its ingredients. The Commission's guidelines on Annex I (Implementing Decision 2013/674/EU) go a step further on stability. Along with the results, the report should include evidence that the composition of the product used for stability testing corresponds to the product placed on the market.
That sentence is the whole argument for doing things in order. Every result is a description of one batch, in one pack, at one point in time. Change the formula, the pack or the process afterwards and the result may no longer describe your product. The tests themselves are only as useful as the decisions you made before you booked them.
Before any sample leaves your workshop: what to freeze
We ask clients to lock four things before the first unit goes to a lab.
- The formula. Exact percentages, and the exact supplier and grade for each ingredient, fragrance included. "Shea butter" from two suppliers can behave differently in a stability oven.
- The primary pack. The container, the closure and any pump, dropper or applicator that touches the product. A dropper that goes back into the bottle after touching skin is a different microbiological risk from an airless pump.
- The process. Order of addition, temperatures, mixing times and the point at which the preservative and fragrance go in. A lab batch made differently from the production batch is a different product as far as stability is concerned.
- The claims list. Every claim you want on the pack and the website. Claims decide which volunteer and efficacy tests you need, and under the common criteria for claims (Regulation (EU) No 655/2013) each one needs evidence that fits it. Our guide to the six common criteria explains how assessors read them.
If one of these is still moving, the honest advice is to wait. A week spent settling the pump is cheaper than a second stability study.
Step zero is a paper review, not a lab booking
The cheapest test is the one your assessor tells you that you don't need, and the most expensive is one run on a formula that has to change anyway. So the first step is sending the frozen formula, the supplier documents, the pack details and the claims list to the safety assessor before you book a lab.
At this stage the assessor is looking for problems that would make any test result useless: an ingredient that is banned or over its limit, a preservative used outside its permitted level or product type, a fragrance whose IFRA limit doesn't fit the product category, or a preservative system that looks thin for an open jar. They also decide which tests are needed. The Commission guidelines split products into three groups for microbiology. Low risk products, such as those with more than 20% alcohol or products at very high or low pH, need neither a challenge test nor microbiological tests on the finished product, though a written justification is still required. Single use products and packs that cannot be opened need microbiological quality testing only. Everything else needs both. ISO 29621 is the standard most assessors use to justify a low risk classification, and our article on water activity and anhydrous products shows how that argument is made.
A quick test of whether you are ready
If you cannot write down the exact formula with percentages, the pack part numbers and the claims list on one page, you are not ready to book a lab. You are ready for a paper review.
Make one representative batch, and make enough of it
Every test in the plan should come from one batch made the way you will make it for sale, from the raw material lots you will actually use, and filled into the final pack. Most of the restarts we see trace back to a sample made in a beaker for one test and a different batch made later for another.
The question we get next is how many units to fill. It varies by lab and by pack size, so always ask the labs for their minimums before you make the batch. As a guide, these are the requests we see most often for a standard 50 ml cream:
| What the units are for | Typical request we see | Notes |
|---|---|---|
| Microbial count at the start (T0) | 1 to 2 units | Taken from the same batch as everything else, before any unit is opened for other work. |
| Challenge test | 1 to 2 units, often 100 to 200 g in total | Ask the lab for its minimum. Small packs such as lip products and 10 ml roll ons may need several units pooled. |
| Stability in the final pack | Roughly 2 units for each storage condition at each time point | Four conditions read at four time points is already about 30 units. Your lab or assessor sets the actual plan. |
| Stability control in inert glass | A smaller matching set | Lets you tell a formula problem from a packaging problem, which the SCCS and the Commission guidelines both mention. |
| Retained samples | 2 to 3 units | Kept from the batch and stored as sold, so you can go back to them if a question comes up later. |
| Volunteer and claims tests (later) | Depends on panel size | Usually filled later from a matching batch, once the assessor is satisfied the product is safe. |
For a 50 ml cream that usually adds up to somewhere between 30 and 40 filled units. Filling five more than you think you need costs almost nothing. Remaking a batch because the lab ran out does.
The order, step by step
Here is the sequence we recommend for a typical water based leave on product. An anhydrous balm or oil will skip some steps, usually the challenge test, if the assessor accepts the low risk justification.
| Step | What happens | Why it sits here | What it hands to the next step |
|---|---|---|---|
| 0. Paper review | The safety assessor reads the full formula, the supplier documents, the pack and your claims list before any lab work. | Catches banned or over-limit ingredients, labelling triggers and preservative gaps while they are free to fix. It also decides which tests you need at all. | A test list, the storage conditions and the questions the lab results must answer. |
| 1. Representative batch | One batch, made with your real process and raw material lots and filled into the final pack, plus a glass set. | Every result afterwards describes this batch. The Commission guidelines ask for evidence that the stability sample matches the product sold. | Enough identical units for every test on the plan. |
| 2. Microbial count (T0) | Total counts and checks for specified organisms against the ISO 17516 limits. | A batch that is dirty at the start makes the challenge result and the stability data hard to rely on. | Confirmation the batch is clean enough to test further. |
| 3. Stability and compatibility start | Units go into storage at the agreed temperatures, plus freeze and thaw cycles where relevant, in the final pack and in glass. | These take longest, so the clock should start as early as possible, on the same day as the challenge test if you can. | Interim readings at agreed time points; later, the shelf life and PAO. |
| 4. Challenge test | The lab adds bacteria, yeast and mould and counts them at 7, 14 and 28 days (ISO 11930). | It needs the frozen formula in its real pH and pack. Run it on the same batch as stability. | Proof that the preservative system works, or a reason to reformulate before anything else is spent. |
| 5. Safety assessment | Part A is compiled and the assessor writes the Part B conclusion, warnings and conditions. | It can start on paper at step 0 and finish once the challenge result and enough stability data are in. | A signed CPSR, label warnings and the durability statement. |
| 6. Volunteer tolerance tests | Patch test, HRIPT or in-use test on people, where you want them. | Tests on volunteers are confirmatory. They should only start once the product has been assessed as safe. | Evidence for tolerance claims such as dermatologically tested. |
| 7. Claims and SPF tests | Efficacy, instrumental or consumer tests, and SPF and UVA testing for sun products. | They must be done on the final formula, so they come after the formula is safe and stable. | Substantiation for every claim on the pack. |
| 8. Label, print and notify | Durability date or PAO, warnings and claims go on the artwork; the product is notified in the SCPN and, for the EU, the CPNP. | Artwork depends on the outputs of steps 3 to 7. | A product you can place on the market. |
A few points that are easy to miss in that table.
The microbial count comes first because it qualifies the batch. The SCCS Notes of Guidance (12th revision) and ISO 17516 set the limits most assessors use: no more than 1,000 colony forming units per gram or millilitre for most products, and no more than 100 for products used around the eyes, on mucous membranes or on children under three, with specified organisms such as Pseudomonas aeruginosa, Staphylococcus aureus and Candida albicans absent. ISO 17516 was reissued as a second edition on 6 November 2026, replacing the 2014 edition; the changes ISO lists are a new normative references clause and an updated bibliography, but ask your lab which edition its report quotes. Our guide to GMP and ISO 22716 covers the hygiene side.
Stability and the challenge test should start together, on the same batch. Stability is the slowest part of the plan, so its clock should start as soon as the batch is cleared. There is no single official protocol: ISO/TR 18811 reviews the published guidance and says it is up to the manufacturer to specify and justify the protocol. What the guidelines are clear about is the pack. The SCCS recommendation quoted in Decision 2013/674/EU says physical stability tests are carried out both in inert containers and in the containers intended for the market. That glass control is what lets you tell a formula problem from a pack problem later. The guidelines also say there are no standard procedures for packaging compatibility, so the assessor relies on knowledge of the materials, supplier data and these stability units in the real pack. Our guide to cosmetic packaging goes into materials.
The challenge test needs the product as sold. ISO 11930:2019 inoculates the product with five reference organisms (Pseudomonas aeruginosa, Staphylococcus aureus, Escherichia coli, Candida albicans and Aspergillus brasiliensis) and counts survivors at 7, 14 and 28 days. Results are judged against criterion A or the less strict criterion B, and criterion B is only acceptable when a risk assessment shows other factors, such as a protective pack, control the risk. A third edition of ISO 11930 was at committee draft stage in 2026, so the 2019 edition with its 2022 amendment is still the reference. Some assessors also ask for a second challenge test on aged stability samples when a preservative is known to lose strength over time or the pH is expected to move. Ask at the paper review rather than finding out at the end.
The assessor does not have to wait for the whole stability study. Many reports are signed on interim stability data with the real time study continuing, as long as the shelf life claimed is one the data supports. Under Article 19(1)(c) a product with a minimum durability of 30 months or less carries a best before date, and one that lasts longer carries a period after opening symbol. The stability data is what justifies whichever applies; our PAO and best before guide explains the choice.
People come after the paperwork, never before it. The Commission guidelines say human volunteer compatibility studies should only be used to confirm safe levels of use, and the scientific committee's guidance on human volunteers in compatibility testing requires the study supervisor to have the full formula and its safety assessment before any test starts. The SCCS Notes of Guidance add that predictive human sensitisation tests should not be carried out. In practice, that means a patch test, a repeat insult patch test (HRIPT) or an in-use study is a confirmation you run once the assessor is satisfied, usually because you want a claim such as "dermatologically tested". It is not a way of finding out whether a formula is safe.
What can run side by side, and what cannot
Overlapping steps is where the time savings are, so it helps to know which overlaps are safe.
- Safe to overlap: the paper review with supplier document chasing; stability, compatibility and the challenge test once the batch is cleared; drafting Part A while the lab work runs; artwork layout while you wait for the warnings and the durability statement; Responsible Person appointment with all of it.
- Not safe to overlap: testing with formula or pack changes; volunteer tests with an unfinished safety assessment; SPF or efficacy tests with a formula that might still change; printing with an unsigned report.
The rule underneath is simple. Anything that describes the product has to wait until the product stops changing. Anything that only uses the description can run alongside.
When a test fails: the restart map
Failures are normal. Roughly speaking, the cost of one depends on how far it sends you back, and that depends on whether the fix changes the formula, the pack or only the process. This is the map we use with clients.
| What failed | Usual cause | What you redo | What you can usually keep |
|---|---|---|---|
| T0 microbial count over the limit | Hygiene, water quality or a contaminated raw material, rather than the formula itself. | Fix the process under your GMP, remake the batch and repeat the count. Start stability and the challenge test again on the clean batch. | The paper review and the formula, if nothing in it changes. |
| Challenge test does not meet criterion A | Too little preservative, the wrong spectrum (mould is a common gap), a pH outside the preservative's working range, or an open pack that lets contamination in. | If the formula changes: a new batch, a new challenge test, a new stability start and an update to the assessment. If only the pack changes to something more protective, the assessor may accept criterion B with a risk assessment. | Supplier documents and the toxicology work on unchanged ingredients. |
| Separation or texture change at high temperature only | Emulsion or thickener system, or a test temperature above anything the product will meet. | Your assessor decides whether a storage instruction and shorter shelf life will do, or whether the formula needs work and a fresh stability start. | The challenge result, if the formula does not change. |
| pH drifts during stability | Missing or weak buffering, an ingredient breaking down, or something leaching from the pack. | Find the cause first. If the pH ends outside the preservative's range, the challenge test may need repeating at the drifted pH or after reformulation. | Microbial count data on the original batch. |
| Pack fails compatibility | Weight loss through the wall, panelling, a corroded pump spring, swelling of a seal, colour change where product meets plastic. | Compatibility and stability in the new pack. A new challenge test only if the old pack was part of the preservation argument. | The formula work, and the glass control data. |
| Reaction in a volunteer test | Irritant level, a sensitiser, or an individual reaction. | Stop. The assessor reviews the case before anything else happens; this is not something to retest quickly. | Nothing should move to launch until the review is done. |
| Fragrance or supplier swapped late | Commercial change after testing started. | Treat it as a formula change. Allergen figures, IFRA check and stability usually need redoing, and the challenge test may too. | Results for parts of the formula the change does not touch, if the assessor agrees. |
Two patterns stand out. First, a formula change almost always restarts stability, because stability describes the formula. Second, a pack change rarely restarts the challenge test unless the pack was part of the preservation argument, as it often is with airless pumps and single dose formats. Whether earlier results carry over to a changed product is decided case by case, and the person deciding should be your assessor, not the person hoping to save a lab fee.
A worked example: a face serum with one failure
Take a 30 ml water based niacinamide serum in a glass dropper bottle, with "suitable for sensitive skin" planned for the label.
- Paper review. The assessor flags that phenoxyethanol at 0.8% on its own is a thin system for a dropper pack, since the pipette touches the face and goes back into the bottle. The formulator adds a second preservative and settles the pH at 5.0 so it sits inside the working range of both. The claim is noted: it will need a volunteer study later, and the CPSR will need to address sensitive skin use.
- Batch. One 10 kg batch, 40 dropper bottles filled, plus a glass set and three retained units.
- Micro count. Well inside 1,000 CFU per gram, specified organisms absent. Cleared.
- Stability, compatibility and challenge test start in the same week. The challenge test passes criterion A for bacteria and yeast but not for Aspergillus brasiliensis.
- The restart. Because the fix changes the formula, the preservative level is raised within its permitted limit, a new batch is made, and both stability and the challenge test start again. Four weeks of stability data on the first batch are kept on file but no longer describe the product.
- Second run. The challenge test meets criterion A. At the agreed interim reading the stability units in glass and in the dropper bottle match, the pH holds at 5.0 and the rubber teat shows no swelling.
- Assessment and claims. The assessor signs Part B with a 12 month best before date supported by the data so far and the real time study continuing. Only then does the serum go to an in-use study on volunteers with sensitive skin to support the claim.
The failure cost about five weeks. Had the formula been changed after a full stability study, it would have cost months.
Where ISO 22716 and ISO 16128 fit
Two standards turn up in test plans that are not tests. ISO 22716:2007 is the good manufacturing practice guideline; under the Regulation, manufacturing in line with the relevant harmonised standards is presumed to meet the GMP requirement, and it governs the batch records, cleaning and raw material controls that make your T0 micro result repeatable. ISO 16128 only calculates natural and organic origin indexes. The standard itself says it does not address human safety or claims, so it belongs in your claims file if you use a "natural origin" percentage, not in your safety testing plan.
Great Britain and the EU: one plan, two reports
The testing logic is identical in both markets. Great Britain applies Regulation 1223/2009 as amended for GB by Schedule 34 of SI 2019/696, and OPSS guidance points to the same Annex I principles, while the EU applies the Regulation directly. A single, well run test plan can support both a GB and an EU safety report, as long as the formula and labels are the same and each report addresses its own market's rules. The differences are in the paperwork around the tests: the Responsible Person, the notification (SCPN for Great Britain, CPNP for the EU) and any rule changes that apply in one market and not the other. Our guide to one CPSR for the UK and EU covers that, and Article 11 sets out what the product information file must hold, test reports included, for ten years after the last batch is placed on the market.
Not sure which tests your product needs, or in what order?
Send us the formula, the pack you plan to use and the claims you want on the label before you book a lab. We will tell you which tests the safety assessment actually needs, the order to run them in, how many units to make, and what we will need back from the lab to sign the CPSR.
Book a launch consultation Packaging compatibility assessment
Related reading
- Do cosmetics need lab testing?
- Cosmetic stability testing explained
- The cosmetic challenge test explained
- Cosmetic microbiological testing explained
- Skin safety and patch testing explained
- How long does a CPSR take?
- How long does a CPNP notification take? The real EU launch timeline
- Cosmetic safety assessment documentation guide
- All regulatory compliance articles
Myths we hear
- A passed challenge test covers any version of the product. It covers the formula, pH and pack that were tested. Whether it carries over to a changed product is the assessor's call, case by case.
- Stability testing can wait until the label is designed. Stability is the slowest part of the plan and decides the durability date or PAO, so it should start as soon as the batch is cleared.
- An HRIPT proves the formula is safe. Volunteer tests only confirm what the safety assessment already concluded. The SCCS says predictive human sensitisation tests should not be done.
- Testing in a lab jar is fine; the pack doesn't matter. The SCCS recommendation in the Commission guidelines is to test in inert containers and in the containers intended for the market.
- The assessor must wait for the whole real time study. Reports are often signed on interim data with the study continuing, provided the shelf life claimed is one the data supports.
What to do this week
- Write the exact formula, supplier list, pack part numbers and claims on one page.
- Send that page and the supplier documents to your assessor for a paper review before booking any lab.
- Ask each lab for its minimum sample size, then plan one batch with five spare units.
- Book the microbial count, stability and challenge test so they start in the same week on that batch.
- Hold volunteer and claims tests until the assessor is satisfied the product is safe.
Frequently asked questions
What order should cosmetic tests be done in?
Paper review by the safety assessor first, then one representative batch in the final pack, a microbial count, then stability, packaging compatibility and the challenge test together, then the safety assessment, and finally volunteer tolerance tests and claims tests.
What should be finalised before cosmetic testing starts?
The formula with exact percentages and suppliers, the primary pack including pumps or droppers, the manufacturing process, and the list of claims. Changing any of them after testing starts can make the results describe a product you no longer sell.
How many samples do I need for cosmetic testing?
It depends on the lab and pack size, so ask each lab for its minimum. For a 50 ml cream, the micro count, challenge test, stability in the final pack, a glass control set and retained samples usually add up to 30 to 40 filled units.
Should stability and challenge tests use the same batch?
Yes, wherever possible. Both results then describe the same product, and the Commission guidelines ask for evidence that the stability sample corresponds to the product placed on the market.
What happens if my challenge test fails?
If the fix changes the formula, you make a new batch and repeat the challenge test and stability, and the safety assessment is updated. If only the pack changes to a more protective one, the assessor may accept ISO 11930 criterion B with a risk assessment.
Do I have to finish stability testing before the CPSR is signed?
Not always. Many assessors sign on interim stability data with the real time study continuing, as long as the best before date or PAO claimed is supported by the data available.
When should an HRIPT or patch test be done?
After the safety assessment has concluded the product is safe. Tests on volunteers are confirmatory, and the SCCS says predictive human sensitisation tests should not be carried out. They are mainly used to support tolerance claims.
Do I need to retest if I change the packaging?
Usually compatibility and stability in the new pack. A new challenge test is needed only if the old pack was part of the preservation argument, for example an airless pump replaced by an open jar.
Official sources and legislation
- Regulation (EC) No 1223/2009 on cosmetic products (EUR-Lex)
- GB Cosmetics Regulation (EC) No 1223/2009 (legislation.gov.uk)
- GB Cosmetics Regulation, Annex I: cosmetic product safety report
- GB Cosmetics Regulation, Article 19: labelling, including date of minimum durability and PAO
- Commission Implementing Decision 2013/674/EU: guidelines on Annex I
- SI 2019/696, Schedule 34: amendments to the Cosmetics Regulation for Great Britain
- Commission Regulation (EU) No 655/2013: common criteria for cosmetic claims
- SCCS Notes of Guidance for the testing of cosmetic ingredients and their safety evaluation, 12th revision (SCCS/1647/22)
- SCCNFP opinion: guidelines on the use of human volunteers in compatibility testing of finished cosmetic products
- ISO 11930:2019: evaluation of the antimicrobial protection of a cosmetic product
- ISO/CD 11930: third edition under development
- ISO 17516:2026: microbiological limits
- ISO 29621:2017: identification of microbiologically low risk products
- ISO/TR 18811:2018: guidelines on the stability testing of cosmetic products
- ISO 22716:2007: cosmetics good manufacturing practices
- ISO 16128-2:2017: natural and organic origin indexes
- OPSS: making cosmetic products available to consumers in Great Britain
Join the conversation