Scale-up is the work of turning a formula that behaves well in a laboratory beaker into the same product made reproducibly in a factory vessel. It matters because a cosmetic emulsion is not just a recipe — it is a recipe plus a process, and the process changes when the batch grows from 500 grams to 500 kilograms. Heat leaves a large vessel far more slowly, shear is applied differently, air is pulled out under vacuum instead of whipped in, and the raw materials arrive as commercial lots rather than samples. The practical consequence is that “it matched the sample” is not a specification. Before you approve a bench sample, agree the numeric ranges — pH, viscosity with its measurement conditions, appearance against a retained physical standard — that the first production batch will actually be judged against.
Key takeaways
- The formula is half the product; the process is the other half. Two batches of an identical formula made with different cool-down rates and different shear histories are two different products in the hand. Nothing in an ingredient list captures that, which is why a formula alone does not transfer cleanly between factories.
- Batch size is set by vessel physics, not by sales policy. A mixer needs its impeller and homogeniser submerged and its scraper sweeping, so most vessels have a minimum working volume of roughly a third to a half of nominal capacity. That is the real floor under a manufacturer’s minimum order quantity.
- A pilot batch is the cheapest insurance in the whole development timeline. Ten to fifty kilograms in production-type equipment surfaces the mixing time, the cool-down curve and the achievable specification range before you have committed to a full order of components.
- “Matches the approved sample” is unenforceable on its own. A sample ages, and the person judging the match changes. A signed, dated physical reference standard held by both parties, plus numeric tolerances, is what makes an out-of-specification batch a fact rather than an argument.
- You will not receive every unit you paid for. Bulk stays on vessel walls, in transfer lines and in the filler hopper. Yield loss and a delivered-quantity tolerance belong in the supply agreement, not in a surprise email at pack-out.
- Small process changes have large sensory consequences. A slower cool-down, a different neutraliser dosing rate or an air-free bulk can move viscosity, colour and slip enough for a customer to notice, while every ingredient remains exactly as approved.
Who this article is for
This is written for brand owners, product managers and sourcing leads who have an approved laboratory sample and are about to authorise a first production run with an OEM (Original Equipment Manufacturer) or ODM (Original Design Manufacturer) partner — and for anyone whose first batch came back subtly wrong and who wants to understand why before the second one. It assumes you are not operating the equipment yourself. If you are still deciding between a stock base and a bespoke formula, start with what custom formulation actually involves and come back here once you have a sample in hand. If you are still choosing the partner, our manufacturer directory is the place to begin.
What is scale-up in cosmetics manufacturing?
Scale-up is the structured process of reproducing a laboratory formula at production volume while keeping the finished product inside an agreed specification. It is a process-engineering exercise, not a formulation one: the ingredient list usually does not change at all. What changes is everything around the ingredients — vessel geometry, heating and cooling rate, the type and duration of mixing and homogenisation, whether the batch is made under vacuum, the order and speed at which materials are dosed, how long the bulk waits before it is filled, and which supplier lots go in.
A separate discipline exists for this because emulsions, gels and suspensions are structured materials. Their properties come from a microscopic architecture built during manufacture — the droplet size distribution of an emulsion, the crystalline network that fatty alcohols and emulsifiers form as the batch cools, the swollen polymer network of a carbomer or xanthan system. That architecture is created by the process. Change the process and you change the architecture, even when the composition is identical to two decimal places.
In an OEM relationship, scale-up sits between sample approval and the first commercial order. It is the stage most often compressed when a launch date is fixed early, and it is the stage where compression is most expensive, because the cost of discovering a problem climbs steeply once components have been ordered and artwork has been printed. If you have not yet reached sample approval, the sequence up to that point is set out in our walk-through of ordering a sample batch from a private-label factory.
Why does the first production batch not match the lab sample?
The production batch differs from the laboratory sample because the physics of a large vessel are genuinely different from the physics of a beaker, and because commercial raw material lots are not sample lots. Both effects are normal, both are predictable, and neither means anyone did anything wrong — but if you have not agreed in advance how much difference is acceptable, both become disputes.
The most under-appreciated factor is heat transfer. For geometrically similar vessels, surface area grows with the square of the linear dimension while volume grows with the cube, so the surface area available per kilogram of product falls as the batch grows. A one-kilogram beaker in a water bath can be taken from 75 °C to 30 °C in well under an hour. A jacketed 500-kilogram vessel doing the same job typically takes several hours, and the exact profile depends on chilled-water temperature, agitator speed and whether the vessel has a wall scraper. That difference in cool-down rate directly changes how the wax and fatty-alcohol network crystallises, which is one of the main things a customer feels as thickness, slip and cushion.
The second factor is shear. Rotational speed alone is a poor comparison between machines: what matters is the shear rate in the rotor–stator gap and how many times the whole batch passes through it. A small laboratory homogeniser at 10,000 rpm on a 25 mm rotor produces a tip speed of roughly 13 metres per second; a production head at 3,000 rpm on a 100 mm rotor produces roughly 16 metres per second. Peak intensity is comparable, but turnover — how much of the batch reaches that zone, and how often — is not. The usual result is a similar mean droplet size with a different distribution width, which shows up as a different viscosity, a different gloss and sometimes different long-term stability.
The table below sets out the parameters that move most often, and what a brand owner typically notices when they do.
| Parameter | In the laboratory | In production | What the brand notices |
|---|---|---|---|
| Heating and cooling rate | Minutes; water bath or hotplate; large surface area per kilogram | Hours; jacketed vessel; heat must leave through a much smaller relative surface | Thicker or thinner than the sample; a waxier, grainier or more “set” texture |
| Shear and homogenisation | High peak intensity, whole batch reached quickly | Comparable tip speed, very different circulation and pass count | Viscosity drift, gloss change, occasionally reduced stability |
| Air content | Air entrained by an open-beaker stirrer | Air removed by vacuum in a modern mixer | Bulk looks denser and slightly darker; fill weight and density change |
| Thermal exposure of actives and fragrance | Short hold at temperature | Longer hold while the batch heats, mixes, cools and waits for release | Softer fragrance top notes; colour drift in extract-containing bases |
| Water | Bench-scale purified water, often a different source | The plant’s own treated water loop, with its own conductivity and microbial specification | pH offset; thickener performance changes with ionic content |
| Dosing rate and order of addition | Poured in by hand over seconds | Metered or pumped over minutes into a moving batch | Local pH and swelling effects; electrolyte-sensitive gels can lose body |
| Raw material lots | Supplier sample lots, often all from one batch | Full commercial lots, including natural-origin materials with genuine lot variation | Colour and odour differences that sit inside every material’s own specification |
| Hold time before filling | Assessed or filled immediately | Bulk may wait hours or overnight for line availability and QC release | Structure continues to build; viscosity read at the wrong moment looks wrong |
None of these effects is exotic. A competent manufacturer expects all of them and designs around them, which is exactly why the batch record — the document capturing actual times, temperatures and speeds rather than intended ones — is worth asking to see. For the wider quality context around that document, our note on quality control practice in cosmetics manufacturing covers where batch records sit in the system, and cosmetic GMP and ISO 22716 explains why they exist at all.
What are the batch stages, and what does each one actually prove?
There are four meaningful stages between an idea and a saleable pallet, and each one answers a different question — which also means each one leaves specific questions unanswered. Confusing them is the root cause of most first-batch disappointments: a bench sample proves the concept, not the manufacture, and people routinely treat it as though it proved both.
| Stage | Typical size | What it proves | What it does not prove | Indicative elapsed time |
|---|---|---|---|---|
| Bench / laboratory sample | 200 g – 2 kg | The concept: sensory direction, colour, fragrance, first-pass compatibility with the active system | Anything about manufacturability, real cost, or behaviour over time | 1–3 weeks per revision round |
| Stability and packaging compatibility set | 1–3 kg, filled into the actual pack | Whether the formula survives time, temperature cycling and the chosen container | Whether it can be made reproducibly at volume | Accelerated read at 4–8 weeks; full programme 12 weeks and beyond |
| Pilot / scale-up trial | 10–50 kg in production-type equipment | The process: mixing and homogenisation times, cool-down curve, achievable specification ranges, realistic yield | Full-line behaviour, long-run fill accuracy, component performance at speed | 2–6 weeks including scheduling |
| First production batch | 200 kg – 1,000 kg and up | Real cost, real yield, real timing, real appearance on shelf | That the next lot reproduces this one — that is what lots two and three prove | 4–10 weeks after materials land |
The stage most often skipped is the pilot. On a genuinely off-the-shelf stock base that the factory has produced many times, skipping it is a defensible commercial decision, because the process is already characterised and the risk sits mostly with your fragrance, colour and pack rather than with the base. On a bespoke formula, or a stock base modified with a new active, a new fragrance load or a substantially different pigment package, skipping the pilot means the first time anyone learns how the formula behaves in a large vessel is the day several hundred kilograms of it exist. The pre-production test panel and stability testing answer different questions again, and neither substitutes for a pilot: a stable formula can still be an unmanufacturable one.
Why minimum working volume decides your minimum order quantity
A production mixer cannot be run part-full. The impeller and the homogeniser head must be submerged, the scraper must be in contact with product, and the vacuum and jacket surfaces need to be working against a real mass. As a rule of thumb, the minimum working volume of a jacketed vessel is somewhere around a third to a half of its nominal capacity, which is why a factory whose smallest suitable vessel is 500 litres cannot sensibly make you 40 kilograms of anything on it.
This is also why a manufacturer’s stated minimum quantity should move when your fill size moves. A 200-kilogram minimum batch fills roughly 4,000 units at 50 g, about 800 units at 250 g and around 400 units at 500 g. A supplier whose minimum is a flat round number of units regardless of fill size is quoting a commercial policy rather than a physical constraint — useful to know, and a legitimate thing to negotiate. We cover the negotiating side in mistakes to avoid when negotiating MOQ, and the underlying definition in what MOQ means in cosmetics manufacturing.
How do you write a specification that makes “matches the sample” enforceable?
You make it enforceable by replacing the sample with two things: a signed, dated physical reference standard held by both parties, and a finished-product specification of numeric ranges with the measurement conditions written next to them. A sample on its own fails as a contractual reference for two reasons — it changes with age, and it can only be compared by subjective judgement, which means whoever is more insistent wins the argument rather than whoever is right.
The reference standard should be produced from the batch you approved, split into sealed retention units, dated, signed by both sides, and stored under stated conditions by both the brand and the manufacturer. When a dispute arises about batch three, you compare batch three against the standard, not against a tube that has been in someone’s handbag since March.
The table below is the specification skeleton we would expect to see on a finished-goods sheet for an emulsion, with the reason each line drifts on scale-up and the objection you should expect when you ask for a tolerance to be written down.
| Specification line | How it must be measured | Why it drifts on scale-up | What a workable tolerance looks like | The objection you should expect |
|---|---|---|---|---|
| Appearance and colour | Visually against the retained standard, under stated lighting, in a stated container | De-aeration, cool-down rate, natural-origin material lots | “Conforms to reference standard; no visible separation, phase or grain” | “Colour varies naturally, we cannot commit to that” |
| Odour | Against the retained standard, by a trained assessor | Thermal exposure during a long hold; fragrance compound lot | “Conforms to reference standard” | “Fragrance is a natural product” |
| pH | Calibrated meter, stated temperature, stated dilution if any | Water ionic content, neutraliser dosing rate, hold time | A band, typically around ±0.3 units on the target | “pH always moves a little” |
| Viscosity | Rotational viscometer with the spindle, speed, temperature, run duration and rest time after manufacture all stated | Cool-down rate and shear history; structure keeps building for days | A range, not a number — for example 18,000–28,000 mPa·s under the stated conditions | “It is not stable until it rests, so we cannot specify it” |
| Relative density | At a stated temperature | Air entrainment versus vacuum de-aeration | A narrow band around the pilot value | “It depends on the filling temperature” |
| Microbiological quality | Total viable count plus absence of specified organisms, by the plant’s validated method | Water loop performance, hygiene, hold time before filling | Category-appropriate limits, tighter for eye-area, mucosal and under-three products | “We only test finished goods on request” |
| Fill weight or volume | Check weights at stated frequency, against the declared content | Bulk density change; viscosity at fill temperature | Per the destination market’s average-quantity rules and an agreed sampling plan | “The filler tolerance is what it is” |
| Delivered quantity | Counted at pack-out against the purchase order | Hold-up in vessel, hoses, pump and hopper; line start-up and changeover losses | An agreed band, commonly a few per cent under to a few per cent over | “Everyone in the industry works to plus or minus ten per cent” |
Two lines on that table deserve emphasis. Viscosity is the most commonly mis-specified property in private-label cosmetics, because a single number with no spindle, speed, temperature or rest time attached is not a measurement at all — two honest laboratories can produce readings that differ by a factor of two from the same jar. And delivered quantity is the line brand owners are most surprised by: it is entirely normal for a run to yield slightly fewer sellable units than the order quantity, and the time to agree what happens then is before the purchase order, not at pack-out. Both belong in the supply agreement rather than in an email; our review of what OEM cosmetics contracts contain covers where these clauses usually sit, and how to read a specification sheet and certificate of analysis covers what you should be receiving with each delivery.
What does scale-up cost, and who pays when a batch falls out of specification?
Scale-up costs money in three places, and only the first is usually quoted to you. The visible cost is the pilot batch itself — materials plus a setup or trial charge, which some manufacturers waive or credit against the first production order if you commit to it. The second, larger and less visible cost is raw material minimums: a fragrance compound may have a supplier minimum of 5 or 25 kilograms, a bespoke pigment dispersion perhaps 20 kilograms, a specialist active a minimum pack size that dwarfs your first batch requirement. Those minimums are often the real reason a quoted MOQ is higher than you expected, and the surplus sits in the factory’s warehouse with its own shelf life attached. The third cost is time: a pilot round adds weeks, and weeks near a launch date have their own commercial price.
Against that, the cost of not doing it is a full production batch that has to be reworked, downgraded or written off. Reworking bulk is sometimes possible — adjusting pH, adding thickener, re-homogenising — and sometimes it is not, particularly once product has been filled into primary packaging. This is why the disposition question matters more than the trial fee.
Before you place the first production order, get four things written down: who decides whether a batch is accepted, reworked or rejected; on what evidence that decision is made; who pays for rework or replacement in each scenario; and who owns the bulk if the filling stage fails after compounding succeeded. A manufacturer willing to put all four in a supply agreement is telling you something useful about how they expect the relationship to run. Note also that if you are buying through an intermediary rather than from the plant itself, these questions get harder to answer, for reasons set out in our comparison of factories, trading companies and sourcing agents. For the underlying cost structure a quotation is built on, see what it costs to manufacture a private-label skincare product.
What should you ask before authorising the first production batch?
Eight questions separate a manufacturer who has thought about your scale-up from one who intends to find out on the day. Ask them in writing, and keep the answers with your product file.
- What is the nominal capacity and the minimum working volume of the vessel this will be made in, and what batch size are you quoting? The answer tells you whether the MOQ is physics or policy, and lets you check the unit arithmetic against your fill size.
- Will you run a pilot batch in production-type equipment before the first production run, what does it cost, and is it credited against the order? If the answer is that it is unnecessary, ask how many times this exact base has been made at this scale.
- What cool-down profile will you use, and will the actual temperature and time be recorded on the batch record? An intended profile and a recorded one are different documents.
- May we see a blank batch record for a comparable product? You are looking for whether it captures actual values with sign-offs, or reads as a printed recipe with nothing filled in.
- What is your purified-water specification, how is the loop maintained, and how often is it tested? Water is the largest ingredient in most emulsions and the one nobody asks about.
- Which raw materials in this formula carry supplier minimums above our batch requirement, and what is the shelf life of the surplus? This surfaces the reorder trap early: a surplus that expires before your second order is a cost you will meet twice.
- What yield loss should we expect between bulk weight and delivered units, and what delivered-quantity tolerance will you commit to? A supplier who has never measured this is telling you something.
- If the batch falls outside specification, who decides accept, rework or reject, and who bears the cost in each case? Ask for the answer in the agreement, not in the email thread.
What goes wrong most often?
Most first-batch failures are process or paperwork failures rather than formulation failures, and the same handful repeat across categories.
- Approving a bench sample with no numeric specification attached. The approval is real, the reference is not, and every later conversation is subjective.
- Letting the reference standard be a photograph or a description. Colour and texture do not survive translation into words or a phone screen. Seal physical retention units, date them, and have both parties sign.
- Going straight from a 500-gram beaker to an 800-kilogram vessel on a novel formula. The jump is roughly three orders of magnitude with no intermediate observation point.
- Comparing a fresh batch against a months-old sample. The sample has aged. Unless it was stored as a controlled retention unit, differences you see may belong to the reference, not the batch.
- Changing the pack after sample approval. A different resin, liner, dispenser or decoration can change compatibility, and the previous compatibility data no longer applies. Our overview of packaging and filling options covers where those interactions arise.
- Measuring viscosity differently at each end. Different spindle, different speed, different rest time, different answer — and an argument that no one can win because neither party is wrong.
- Accepting “within our normal tolerance” where no tolerance was ever agreed. If it is not in the specification, it is not a tolerance; it is a preference.
Once the first batch is right, the discipline that keeps it right is reproducibility across lots — which is a different problem, and one we cover in reordering and scaling production. Batch identification and retention sampling, covered in shelf life, PAO and batch codes, is what lets you investigate a complaint on lot four by comparing it with what you actually shipped.
Frequently asked questions
How long after manufacture should viscosity be measured?
At the rest time written into the specification, and at the same rest time by both parties. Emulsions and polymer-thickened gels continue to build structure after manufacture, often for 24 to 72 hours and sometimes longer, so a reading taken straight off the vessel and one taken after three days can differ substantially for a perfectly good batch. Agree a single rest time, a single spindle and speed, and a single measurement temperature, and put all four on the specification sheet next to the range.
Is a pilot batch worth the cost on a stock formula?
Often not, if the base is genuinely unmodified and the factory makes it routinely at the scale you are ordering. The calculation changes as soon as you alter anything — a new fragrance at a meaningful dose, a new active, a pigment change, or a switch of pack that affects filling. Ask how many times this exact base, at this exact scale, has been produced in the past year; if the answer is a small number or a hesitation, the pilot has just become cheap relative to the risk.
Why did I receive fewer units than I ordered?
Because some of the bulk never becomes a saleable unit. Product remains on vessel walls and the agitator, in the transfer hose and pump, and in the filler hopper, and more is lost to line start-up and end-of-run adjustment. The proportion is larger for viscous products, smaller batches and longer transfer paths. This is normal and should be handled by an agreed delivered-quantity tolerance in the supply agreement, with a stated rule for what happens if the shortfall exceeds it.
Can a different factory reproduce my formula exactly?
It can reproduce your formula; reproducing your product is harder. A new manufacturer has different vessel geometry, different homogenisers, a different cool-down capability, a different water system and different raw material suppliers, so a straight formula handover typically lands close but not identical, and closing the gap takes a matching exercise against a retained standard. Budget for at least one pilot round and a fresh stability and compatibility programme when you move a product, and keep the ownership of the formula unambiguous before you start — see protecting your formula and IP with an OEM manufacturer.
My batch smells slightly different from the sample. Is it out of specification?
Only the specification can answer that, which is the point. Fragrance compounds are supplied to their own tolerance, natural-origin materials vary by lot, and a longer hold at temperature in a large vessel takes some edge off the top notes even when nothing is wrong. Compare the batch with the sealed reference standard rather than with an opened sample, ask the manufacturer for the fragrance certificate of analysis for the lot used, and if the difference is real and outside what you agreed, raise it as a formal deviation while the batch is still unfilled.
Sources, scope and limitations
General cosmetic manufacturing obligations in the European Union, including good manufacturing practice and the product information file that a batch record ultimately feeds, are set out through the European Commission’s cosmetics legislation portal, with ingredient status searchable in CosIng and safety opinions published by the Scientific Committee on Consumer Safety. The equivalent Great Britain requirements sit in the Cosmetic Products Enforcement Regulations 2013. In the United States, the Food and Drug Administration publishes draft guidance on cosmetic good manufacturing practices, guidance on microbiological safety in cosmetics, the Cosmetics Labeling Guide, and pages covering facility registration and product listing and compliance and enforcement under the Modernization of Cosmetics Regulation Act. ASEAN product information file expectations are summarised by Singapore’s Health Sciences Authority under the ASEAN Cosmetic Directive. Good manufacturing practice for cosmetics is described by ISO 22716, preservation efficacy testing by ISO 11930, microbiological risk assessment by ISO 29621, and microbiological limits for finished cosmetics by ISO 17516 — the standards against which most of the certificates and test reports discussed here are issued.
Limitations. The batch sizes, vessel working-volume proportions, cool-down times, tip-speed arithmetic, elapsed times and tolerance bands given here are illustrative engineering rules of thumb and ordinary commercial practice observed in private-label cosmetics projects. They are not surveyed data, not regulatory requirements, and not a substitute for your own manufacturer’s process data. Real values depend on vessel geometry, jacket and chiller capacity, formula chemistry, viscosity, transfer distance and fill format, and vary widely between plants. Microbiological limits, average-quantity rules and documentation obligations differ by market and product category. We did not audit any manufacturer or observe any specific plant in preparing this article. Confirm your own position against the primary sources linked above, against your supplier’s own validated methods, or with a qualified regulatory adviser in your destination market, before approving a specification or authorising a production batch. Last reviewed 8 September 2026.
This article is general information for brand owners and product teams, not formulation, engineering, legal or regulatory advice. Manufacturing equipment, process capability, specification practice and market requirements differ by factory, category and jurisdiction and change over time — confirm your own position with your manufacturer and a qualified adviser before approving a specification or releasing a production batch.