Delta E (ΔE) is a single number that expresses how far apart two colours are in the CIE L*a*b* colour space, and it is the standard way to write a colour tolerance into a cosmetics manufacturing agreement. A ΔE near 1 is close to the threshold a trained eye can see; most commercial tolerances for cosmetic packaging and tinted formulas fall between about 1 and 3, depending on the component and the formula used. This guide covers packaging colour, product colour and the tolerance clause for brands sourcing from OEM or ODM factories in any market. The number is meaningless unless its formula, illuminant, instrument and sample preparation are stated beside it. Ask which formula your factory reports first.
Key takeaways
- ΔE is a distance, not a grade. It measures how far a batch sits from the approved standard in a three-axis colour space (lightness, red–green, yellow–blue). A ΔE of 2.0 made entirely of a lightness shift looks different from a ΔE of 2.0 made of a hue shift, which is why the components matter as much as the total.
- The formula changes the number. The same red lipstick pair measures ΔE*ab 3.08 under the 1976 formula and ΔE00 1.18 under CIEDE2000 in our worked example. A tolerance written without naming its formula cannot be enforced.
- Packaging and formula need different tolerances. Illustrative starting points: ΔE00 1.0–1.5 for an opaque moulded cap against its colour plaque, about 2.0 for a spot colour on a printed carton, 1.0–1.5 for a foundation bulk against its sealed standard, and a b* limit rather than a ΔE limit for white creams that yellow over shelf life.
- Four things must be fixed in writing with the number: the colour-difference formula, the illuminant and observer, the instrument geometry, and how the sample is prepared. Change any one and the same batch produces a different result.
- The instrument settles the argument; the light booth decides acceptability. A batch can pass its ΔE limit and still look wrong because of metamerism or gloss. A workable agreement gives the number and a multi-illuminant visual assessment each a defined role.
Who is this article for?
This explainer is for brand owners, product managers and sourcing staff who buy finished cosmetics or packaging components from an OEM (Original Equipment Manufacturer) or ODM (Original Design Manufacturer) and need to approve colour, reject colour, or write a tolerance into a quotation, specification or supply contract. It assumes no colour-science background. If your immediate problem is a makeup range with many shades, our detailed guide to developing a makeup shade range with an OEM manufacturer covers shade architecture, approval rounds and drydown; this article explains the number that guide tells you to put in the contract. If you are still choosing a factory, the guide to choosing an OEM cosmetics manufacturer and the OEMHallmark manufacturer directory are the places to start.
What is Delta E in plain terms?
Delta E is the straight-line distance between two colours plotted in the CIE L*a*b* colour space, so a ΔE of zero means an identical measurement and larger values mean a larger difference. The International Commission on Illumination (CIE, from its French name) defined the L*a*b* space in 1976 so that equal distances would correspond, approximately, to equally visible differences; the current definition is published as ISO/CIE 11664-4, Colorimetry Part 4: CIE 1976 L*a*b* colour space (CIE publication page, checked 30 October 2026). Three coordinates locate a colour: L* for lightness (0 = black, 100 = white), a* for the red–green axis (positive = redder) and b* for the yellow–blue axis (positive = yellower). A warm beige foundation might read L* 62.0, a* 18.5, b* 21.0; a navy cap might read L* 35.2, a* −2.1, b* −38.5.
The original 1976 colour difference, written ΔE*ab, is simply the three-dimensional distance between the two points: the square root of (ΔL*)² + (Δa*)² + (Δb*)². If a batch of that foundation measures L* 61.2, a* 18.9, b* 21.6, the differences are −0.8, +0.4 and +0.6, and ΔE*ab is the square root of 0.64 + 0.16 + 0.36, which is 1.08. Instrument makers explain the same geometry in their own training material; the Konica Minolta Precise Color Communication guide to colour spaces and the X-Rite article Tolerancing Part 3: Color Space vs. Color Tolerance (both checked 30 October 2026) are the two most-cited introductions and are free to read.
How visible a given ΔE is depends on the colour, the observer, the lighting and whether the two samples sit side by side. The bands below are the rule of thumb most colour laboratories work to for ΔE*ab under a light booth. They are a convention, not a standard, and the shade-range guide linked above already warns that commercial cosmetic tolerances are usually looser than the threshold of perception.
| ΔE*ab band | What a trained observer sees under a light booth, samples adjacent | How it is usually treated in a cosmetics supply chain |
|---|---|---|
| Below 1.0 | Not normally distinguishable; differences at this level are also close to the agreement between two good instruments | Instrument-level match; the tightest tolerance a factory will realistically accept for a tinted bulk |
| 1.0 to 2.0 | Visible to a trained observer on close side-by-side comparison; rarely noticed by a consumer | Typical tight commercial tolerance for complexion products and opaque moulded components |
| 2.0 to 3.5 | Noticed at a glance by an untrained observer when the samples are adjacent | Common for printed packaging and decorated components; acceptable where the consumer never sees two batches together |
| 3.5 to 5.0 | Obvious difference | Usually a reject for anything that touches the brand colour; sometimes tolerated end-of-shelf-life for product colour |
| Above 5.0 | Reads as a different colour | Reject |
Two cautions follow from that table. First, a ΔE is only meaningful between a sample and a defined reference, so the first task in any colour programme is agreeing what the reference is. Second, the bands shift with the formula used, as the next section shows, so a tolerance that reads “ΔE 2” with no formula stated is an invitation to argue.
Which Delta E formula should you write into a cosmetics specification?
Specify CIEDE2000 (written ΔE00) unless your factory or component supplier already reports a different formula consistently, in which case specify that formula by name and never mix tolerance values between formulas. The 1976 distance treats every direction and every region of colour space equally, and that is not how people see: a small shift in a saturated red is far less visible than the same numeric shift in a pale beige. The CIE published CIEDE2000 to correct for this, with weighting for lightness, chroma, hue and a chroma–hue interaction in the blue region; it is standardised as ISO/CIE 11664-6, Colorimetry Part 6: CIEDE2000 colour-difference formula (checked 30 October 2026). ASTM’s practice D2244-23, Calculation of Color Tolerances and Color Differences from Instrumentally Measured Color Coordinates (checked 30 October 2026) lists CIELAB, CMC, CIE94, DIN99o and CIEDE2000 as the recognised systems and states plainly that purchaser and seller must agree the tolerance and the calculation procedure, and that differences obtained in one system should only be compared with differences in the same system.
| Formula | Origin | What it does | Where you will meet it | Limitation for a cosmetics brand |
|---|---|---|---|---|
| ΔE*ab (CIE 1976) | CIE, 1976; ISO/CIE 11664-4 | Straight-line distance in L*a*b* | Default output of older instruments and many factory certificates of analysis; older packaging specifications | Over-reports differences in saturated colours and under-reports them in pale ones; a tolerance that fits a beige does not fit a red |
| ΔE*94 (CIE94) | CIE, 1995 | Divides the chroma and hue terms by factors that grow with chroma | Some plastics and coatings specifications | Result depends on which sample is called the standard; largely superseded by CIEDE2000 |
| ΔE CMC(l:c) | Colour Measurement Committee of the Society of Dyers and Colourists, 1984; usually written CMC(2:1) for acceptability | Defines an ellipsoidal tolerance volume around the standard, with lightness weighted by the l:c ratio | Textiles, plastics masterbatch, some label printers | Not symmetric; the l:c ratio must be stated or the number is undefined |
| ΔE00 (CIEDE2000) | CIE, 2001; ISO/CIE 11664-6:2014 | Corrects the 1976 distance for lightness, chroma, hue and the blue-region chroma–hue interaction | Recommended current formula; supported by every modern spectrophotometer; included in ASTM D2244 | Numbers are generally smaller than ΔE*ab for the same pair, so a tolerance copied from an old ΔE*ab specification becomes much looser if the formula is switched without re-setting the limit |
The practical consequence is easiest to see on real pairs. The table below takes four illustrative standard-versus-batch pairs, calculated by us with the published CIEDE2000 procedure and checked against the CIE reference test data, and shows how far the two formulas disagree. None of these is a measurement from a named factory; they are chosen to show the arithmetic.
| Illustrative pair | Standard (L*, a*, b*) | Batch (L*, a*, b*) | ΔE*ab (1976) | ΔE00 (CIEDE2000) | What the gap tells you |
|---|---|---|---|---|---|
| Warm beige foundation, batch slightly darker and yellower | 62.0, 18.5, 21.0 | 60.3, 18.4, 21.9 | 1.93 | 1.59 | Moderate agreement; the lightness shift dominates both numbers |
| Navy cap, batch slightly lighter and less blue | 35.2, −2.1, −38.5 | 34.0, −0.9, −36.8 | 2.40 | 1.68 | CIEDE2000 discounts the chroma change in a saturated blue |
| White cream base yellowing on stability (b* 2.0 to 4.5) | 94.5, −0.6, 2.0 | 94.2, −0.8, 4.5 | 2.53 | 2.20 | Close agreement in near-whites; both formulas see yellowing clearly |
| Saturated red lipstick, batch marginally darker and less chromatic | 42.0, 58.0, 30.0 | 41.0, 55.5, 28.5 | 3.08 | 1.18 | Large disagreement; a ΔE*ab 2.0 limit rejects this batch and a ΔE00 2.0 limit accepts it comfortably |
The red lipstick row is the one to remember. A factory that reports ΔE*ab and a brand that reads the result as if it were ΔE00, or the reverse, will reject and accept the wrong batches. The same confusion appears in reverse when a brand sets a CIEDE2000 limit of 1.0 because it read somewhere that “ΔE 1 is invisible”; that rule of thumb was written for the 1976 formula, and 1.0 ΔE00 on a saturated shade is tighter than many pigment dispersions can hold batch to batch.
Direction matters as well as size. Three batches of the same beige foundation can each sit at exactly ΔE*ab 2.0 from the standard, one by being 2.0 units darker (L* 60.0), one by being 2.0 units redder (a* 20.5) and one by being 2.0 units yellower (b* 23.0). Under CIEDE2000 those three measure 1.73, 1.34 and 1.22 respectively, because a lightness error at this chroma is more visible than a chroma error of the same size. That is why a careful specification pairs the total ΔE limit with a separate limit on ΔL*, which is the component consumers notice first on a complexion product.
How tight should a colour tolerance be for cosmetic packaging and formulas?
There is no universal cosmetics tolerance; the right limit depends on the component, the material, how the consumer will see it and what the process can hold, so the table below gives illustrative starting points that a brand can then tighten or loosen with its factory. Two principles set the range. A tolerance tighter than the process capability guarantees rejected batches and a factory that stops measuring honestly; a tolerance looser than what a consumer notices on shelf, especially where two components of the same brand colour sit next to each other, guarantees complaints. Where a cap, a bottle and a carton all carry the brand colour in different materials, the realistic goal is that each sits close to a single master so that no pair in the set is more than roughly ΔE00 2 apart; the pair-to-pair difference is what the shopper sees.
| Component or product | What is compared with what | Illustrative starting tolerance | Instrument geometry usually used | Sample preparation that must be fixed | Who normally measures |
|---|---|---|---|---|---|
| Opaque moulded plastic cap, bottle or jar (masterbatch coloured) | Moulded part versus approved colour plaque or masterbatch standard | ΔE00 1.0–1.5 for smooth gloss parts; up to 2.0 for textured or matte finishes | d/8 sphere with specular included (SCI) to remove gloss, or 45/0 | Flat area of the part, or a plaque moulded from the same resin and masterbatch lot at the same thickness | Component supplier at release; factory at incoming inspection |
| Printed label or folding carton, solid spot colour | Press sheet versus approved printed proof on the same substrate | ΔE00 about 2.0 for a brand spot colour; 2.5–3.0 for process-colour areas | 45/0 spectrophotometer on the solid patch | Measured on a solid patch of the same ink film on the same stock; white backing stated | Printer in-line or at sheet pull; brand or factory on a sample at incoming |
| Screen print, hot foil or pad print on a bottle or tube | Decorated part versus approved decorated standard | ΔE00 2.0–2.5, often with a visual-only clause for small or curved areas | 45/0 with a small aperture where a flat area exists; otherwise visual under booth | Decoration measured over the same base material colour; curved parts assessed visually | Decorator; factory at incoming |
| Foundation, tinted moisturiser, BB or CC cream (bulk) | Bulk batch versus sealed physical standard | ΔE00 1.0–1.5, plus ΔL* within about ±0.7 | 45/0 or d/8 with specular excluded (SCE) | Drawdown at a fixed wet-film thickness on an opacity card, dried for a stated time; measured over the white and black halves if hiding is in scope | Factory laboratory at every batch; reported on the certificate of analysis |
| Lipstick, pressed powder, cream blush | Batch versus sealed standard | ΔE00 1.5–2.0 | d/8 SCE or 45/0 | Lipstick: smear at fixed pressure on a stated card; powder: pressed puck at stated pressure, or a loose fill in a cell | Factory laboratory |
| White or pale emulsion (cream, lotion, cleanser) | Batch versus standard at release; batch versus itself across shelf life | b* within ±1.0 of standard and ΔE00 1.5 at release; a separate end-of-shelf-life limit of about ΔE00 3 on stability | d/8 SCI | Filled into a fixed-depth cell or dish to a stated depth; measured at a stated temperature | Factory laboratory at release; stability laboratory at pull points |
| Clear or translucent liquid (toner, shampoo, oil, micellar water) | Batch versus standard in a fixed path-length cell | Gardner or APHA/Pt-Co colour-scale limit, or transmission ΔE00 about 2 in a 10 mm cell | Transmission spectrophotometer or visual against reference solutions | Fixed cell path length; bubble-free; stated temperature | Factory laboratory |
Every figure in that table is a starting point for negotiation, not a standard; the factory’s own batch history is the only valid basis for the final number. A useful way to find it is to ask for the last ten batches’ ΔE values against the standard before you set the limit. If nine of ten sit under 1.2 and one sits at 1.8, a tolerance of 1.5 is honest and a tolerance of 1.0 is a fight waiting to happen. Our tutorial on reading a cosmetic spec sheet and certificate of analysis shows where colour normally appears on those documents and what a result reported as “conforms” hides.
Packaging deserves one more note. Print standards for offset lithography, principally ISO 12647-2, set tolerances for process-colour solids that are written for printers reproducing images, and many brands find them too loose for a brand colour. A spot colour mixed to a Pantone reference is the usual answer; the Pantone colour systems page (checked 30 October 2026) explains that the swatches are specified ink on coated or uncoated paper, which is why the same number on a white polypropylene label, a clear film and a kraft carton gives three different results. The physical printed proof, archived as the standard, is what a ΔE is measured against; our step-by-step on getting cosmetic packaging artwork print-ready explains how that proof is produced and approved.
How is colour actually measured on a cosmetic product or pack?
Colour is measured with a spectrophotometer or colorimeter under a stated illuminant and observer, from a sample prepared in a stated way, and the result is then compared with the standard measured on the same instrument in the same way; change any of those conditions and the ΔE changes. The elements that have to be fixed are listed below, because each one is a place where a factory and a brand can both be measuring correctly and still get different numbers.
Illuminant and observer. The L*a*b* coordinates of a sample depend on the light it is assumed to be seen under and on the standard observer used to convert the spectrum to numbers. Cosmetics and packaging normally specify illuminant D65 (average daylight) with the CIE 1964 10° observer; printing often uses D50 with the 2° observer. The CIE defines both observers in Colorimetry Part 1 and the illuminants in Colorimetry Part 2 (BSI product page; both checked 30 October 2026); the Konica Minolta guide to the 2° and 10° standard observers explains the difference in plain language. The factory certificate must state which pair was used.
Instrument and geometry. A tristimulus colorimeter reads L*a*b* through three filters under one fixed illuminant; a spectrophotometer records the full reflectance spectrum and can calculate colour under any illuminant, which is what makes metamerism detectable. Geometry describes how the sample is lit and viewed. A 45/0 instrument lights at 45 degrees and views from the perpendicular, which excludes gloss and corresponds to how a person looks at a printed label. A d/8 sphere instrument lights diffusely and views at 8 degrees; with the specular component included (SCI) it ignores gloss and texture and reports the colour of the material, with it excluded (SCE) it reports what the eye sees including gloss. Caps and bottles are usually specified SCI so that a gloss variation does not read as a colour variation; a drawdown of foundation is usually SCE because gloss is part of what the consumer sees. ASTM’s test method E1347 for colour and colour-difference measurement by tristimulus colorimetry (checked 30 October 2026) sets out the instrument conditions that have to be matched for two readings to be comparable.
| Method | What it reports | Best used for | Weak point | Who usually has one |
|---|---|---|---|---|
| Light booth with trained observers (D65, store fluorescent, incandescent A) | Pass, fail or borderline by eye; detects metamerism when sources are switched | Final commercial acceptability; curved, textured and small decorated parts; complexion products on skin | No number; observers tire and disagree; booth lamps age and must be replaced on schedule | Every cosmetics factory and decorator should; many brands do not |
| Tristimulus colorimeter | L*a*b* and ΔE under one built-in illuminant and observer | Fast batch-to-batch checks of flat opaque samples against a standard measured on the same unit | Cannot detect metamerism; agreement between two different colorimeters is poor, so brand and factory cannot compare numbers | Smaller factories and printers |
| 45/0 spectrophotometer | Reflectance spectrum; L*a*b* and ΔE under any illuminant | Labels, cartons, drawdowns, pressed powders; anything flat | Gloss and texture change the reading, so sample preparation must be identical | Printers, colour laboratories, better-equipped factories |
| d/8 sphere spectrophotometer (SCI and SCE) | As above, with gloss included or excluded by setting | Moulded plastics, caps, bottles, powders; comparing a matte part with a gloss standard | SCI can pass a part whose gloss difference the eye rejects; the setting must be written down | Masterbatch and component suppliers; larger cosmetics factories |
| Transmission spectrophotometer or Gardner/APHA comparator | Colour of a liquid in a fixed cell | Clear shampoos, toners, oils, fragrances | Cell path length, bubbles and temperature all change the result | Factory laboratories running liquid products |
Sample preparation. This is where most colour disputes actually live. A foundation measured as a 100-micron drawdown reads differently from the same batch measured as a 200-micron drawdown, in a dish, or on skin after drydown; a pressed powder measured on its puck reads differently from a loose fill; a moulded cap measured on its curved shoulder reads differently from the flat top. The specification has to name one preparation, and both laboratories have to use it. The shade-range guide linked earlier already covers drydown for complexion products; for packaging the rule is a plaque or a flat area of the part, measured over a stated backing.
Metamerism. Two samples can match under D65 and visibly differ under a store’s fluorescent lamps because they reach the same colour through different pigment or ink combinations. An instrument reports it as a different ΔE under each illuminant; a booth shows it when the lamp is switched. It is not a defect a factory can adjust away on request, because it is a property of the colourant blend chosen at development, so it has to be caught at approval by assessing every submission under at least D65, a store-type fluorescent source and an incandescent source. ASTM’s practice D1729-22 for visual appraisal of colours and colour differences (checked 30 October 2026) specifies booth conditions, recommends D65 as the primary source and D50 for printing applications, and is the natural reference to cite for the visual half of a clause. Where a colour will be assessed on skin, the booth is also where the brand learns that the batch within tolerance on a drawdown is not within tolerance on a face.
Colourant permission is separate from colour measurement. A ΔE clause says nothing about whether the pigments are allowed. In the United States, colour additives permitted in cosmetics are listed in 21 CFR Part 73 Subpart C (exempt from certification) and 21 CFR Part 74 Subpart C (subject to batch certification), summarised on the FDA colour additive summary table; in the European Union, permitted colourants are Annex IV of Regulation (EC) No 1223/2009, accessible through the Commission’s cosmetics legislation page and searchable in the CosIng database; Great Britain applies its own retained version described in the UK government guidance on making cosmetic products available in Great Britain; and ASEAN markets follow the annexes of the ASEAN Cosmetic Directive, explained on the Health Sciences Authority of Singapore’s ACD page (all checked 30 October 2026). A re-shade that swaps a pigment to hit a tolerance can take a product out of compliance in one market while fixing its colour in another.
What should a colour-tolerance clause in an OEM agreement contain?
A workable clause names the physical standard, the formula, the numeric limit and its components, the illuminant and observer, the instrument geometry and aperture, the sample preparation, the sampling plan, what the certificate must report, and what happens when the number passes but the eye disagrees. Most quotations and many signed agreements contain only the phrase “colour to approved standard” or “within industry tolerance”, neither of which can be tested. Our review of examples of OEM cosmetics contracts shows where a quality schedule normally sits in the document; the table below is what the colour part of that schedule should contain, with example wording a brand can adapt.
| Clause element | Why it has to be written down | Example wording (adapt to your product) |
|---|---|---|
| Reference standard | A ΔE has no meaning without a defined reference; memory and photographs are not references | “The colour standard is sealed physical master No. 2026-FD-03, approved on [date], one copy held by each party, stored below 25 °C in the dark, re-issued jointly every 12 months or when the master’s own measurement drifts more than ΔE00 0.5 from its stored spectral data.” |
| Formula and units | The same pair gives different numbers under different formulas | “Colour difference is calculated as CIEDE2000 (ΔE00) per ISO/CIE 11664-6 with kL:kC:kH = 1:1:1.” |
| Limit and components | A total can hide a lightness shift that consumers notice first | “Each batch shall measure ΔE00 ≤ 1.2 against the standard, and ΔL* within ±0.7.” |
| Illuminant and observer | Coordinates change with the assumed light source | “Calculated for illuminant D65 and the CIE 1964 10° observer; visual assessment under D65, a store-type fluorescent source and illuminant A in a booth meeting ASTM D1729.” |
| Instrument and geometry | 45/0 and d/8, SCI and SCE, and aperture size all shift the result | “Measured on a d/8 spectrophotometer, specular excluded, 8 mm aperture, calibrated daily against its white tile; brand and factory instruments to agree within ΔE00 0.3 on a shared set of ceramic tiles, checked quarterly.” |
| Sample preparation | The single largest source of unexplained disagreement | “Bulk drawn down at 100 µm wet film on a Leneta-type opacity chart, dried 24 hours at 23 °C, measured over the white half; three readings averaged per sample.” For a cap: “a 2 mm plaque moulded from the same resin and masterbatch lot, measured on the flat face.” |
| Sampling | One reading from one jar is not a batch result | “Bulk: top, middle and bottom of the batch vessel after final mixing. Components: sampled per ISO 2859-1 at the inspection level stated in the component specification.” |
| Reporting | “Conforms” on a certificate cannot be audited | “The certificate of analysis shall state L*, a*, b*, ΔE00 and ΔL* against the standard, the instrument serial number and the measurement date.” |
| Visual override | A batch can pass numerically and fail by eye through metamerism or gloss | “Numeric conformance does not of itself constitute acceptance. A batch within tolerance that two trained observers reject under booth assessment is referred to joint review within five working days; a third-party colour laboratory’s measurement on the agreed preparation is final on the numeric question.” |
| Consequence | Without one, the tolerance is advisory | “Batches outside tolerance are reworked or replaced at the manufacturer’s cost, with the delivery date extended by the rework time only.” |
Two refinements make the clause fairer and therefore more likely to be signed. First, grade the limit by stage: a tighter number at first production, when the dispersion lot and the process are fresh, and a documented process for re-setting it once ten batches of history exist. Second, accept that a component supplier two tiers down will have its own tolerance with the factory, and ask to see it; a cap specified at ΔE00 1.0 by the brand but bought by the factory at ΔE00 2.5 from the moulder is a dispute the brand will lose. Our checklist of documents to request from a cosmetics manufacturer includes the component specifications for this reason.
Where does colour drift between approval and the tenth batch?
Colour drifts for physical reasons that are mostly predictable: a new pigment or masterbatch lot, a change in process heat or shear, ageing of the standard, substrate variation in print, and chemical change in the product over its shelf life. Each has a characteristic signature, and knowing it turns an argument about who is at fault into a diagnosis.
| Cause of drift | Where it shows | Typical signature | Prevention that belongs in the specification |
|---|---|---|---|
| New pigment dispersion or masterbatch lot | Tinted bulk; moulded components | Step change between two batches, then stable | Lot-to-lot ΔE limit on the dispersion or masterbatch itself; retain samples of each lot |
| Process heat, hold time or shear at production scale | Emulsions with extracts or natural colourants; pigmented bulks | Batch darker or yellower than the laboratory sample it was approved against | Approve the standard from a pilot or first production batch, not a bench sample; our guide to scale-up in cosmetics manufacturing explains why bench and plant batches differ |
| Base formula change after shading | Colour cosmetics | Every shade moves at once | Lock the vehicle before shading; any base change re-opens shade approval |
| Ageing of the physical standard | Everything compared with it | Slow, one-directional drift of batches that are actually stable | Store the spectral data of the master at approval; re-measure the master before every dispute |
| Substrate or ink lot change in print | Labels, cartons, printed tubes | Whole reprint shifted; proof still matches the first run | Specify stock brand and grade; require a new proof when either changes |
| Regrind or recycled content in moulding | Caps, bottles, jars | Greyer, duller or slightly darker parts with the same masterbatch | State the permitted regrind or PCR percentage; measure SCI and SCE so gloss loss is separated from colour |
| Chemical change over shelf life | White and pale emulsions; products with vitamin C, retinoids, natural extracts, some fragrances | Yellowing or browning that accelerates with heat and light | Set an end-of-shelf-life colour limit in the stability protocol, not just a release limit; see the cosmetic product testing checklist for the pull points at which colour is recorded |
Colour also has to survive the factory’s own gate. A component that arrives outside tolerance should be caught at incoming inspection, not on the filling line, and the inspection plan has to say what instrument and standard the inspector uses; our tutorial on incoming inspection and AQL sampling for cosmetic packaging covers the sampling arithmetic and the cost of finding a colour defect at each later point.
How do you set up a colour programme with a new factory?
A brand with no colour laboratory can still run a disciplined programme by owning the standard, specifying the method and buying one shared instrument or one third-party measurement per batch. The steps below are the sequence that works for a first product and scales to a range.
- Decide which colours matter commercially. A brand colour that appears on cap, bottle and carton; a complexion shade; a white base that must not yellow. Everything else can carry a looser visual-only tolerance.
- Approve against physical samples under a booth, never a screen. Assess under D65, a store fluorescent source and incandescent light, after drydown for products applied to skin.
- Seal the standard and store its spectral data. Two copies, dated and signed; the factory measures it on the agreed instrument on day one and files the spectrum.
- Ask the factory for ten batches of history on a comparable product. Set the limit from that distribution, with the formula named.
- Write the clause using the ten elements in the table above, and attach it to the quotation before the first production order, not after the first complaint. If the factory is a trading company rather than the manufacturer, make sure the clause reaches the plant that measures; our comparison of a factory, trading company and sourcing agent explains why that matters.
- Require numbers on every certificate (L*, a*, b*, ΔE00, ΔL*, instrument, date), and plot them. A trend toward the limit is visible several batches before a failure.
- Retain a sample from every batch in the same conditions as the standard, so that a complaint can be checked against the batch rather than against the master.
- Re-set the limit after ten batches, tighter if the process allows it and the product is colour-critical, looser if the first number was aspirational.
Questions that reveal quickly whether a factory can hold colour: Which colour-difference formula does your laboratory report, and can you report CIEDE2000? What instrument and geometry do you use, and when was it last verified against ceramic tiles? What sample preparation do you use for this product type? Can you show the ΔE history for an existing product in this category? Who holds the colour standards for your current clients, and how are they re-issued? A factory that answers all five with specifics is unusual and worth keeping; a factory that answers “we match to the standard” is not lying, but it is not measuring either.
The mistakes that recur: approving on a screen, a phone photograph or a Pantone chip; setting ΔE 1 because a blog said it was invisible, then rejecting half the batches; copying a ΔE*ab limit into a CIEDE2000 clause unchanged; forgetting that SCI and SCE readings of the same cap differ; writing a release limit for a white cream with no shelf-life limit; and letting the factory hold the only copy of the standard. Each is avoidable with a line in the specification.
Frequently asked questions
What Delta E value is considered acceptable for cosmetics?
There is no single accepted value, because acceptability depends on the product, the colour and the formula used. As illustrative starting points, complexion products and opaque moulded components are often held to CIEDE2000 1.0 to 1.5 against a sealed standard, printed packaging to about 2.0 for a brand spot colour, and white bases to a separate yellowness (b*) limit. Ask your factory for ten batches of measured history before setting any number, and state the formula beside it.
Is Delta E 1 really invisible?
Roughly, and only under conditions that rarely apply on a shelf: a trained observer, a controlled light booth, samples side by side, and the 1976 ΔE*ab formula. The figure is a widely used rule of thumb rather than a standard, perceptibility differs by colour region, and 1.0 under CIEDE2000 is tighter than 1.0 under the 1976 formula for most saturated colours. Treat it as a reference point, not a specification.
Can I use a Pantone number as my colour tolerance?
A Pantone number identifies a target, not a tolerance, and the swatch is ink on a specific paper. Use it to communicate the intent, then convert it into a physical standard on your real material (a printed proof on the actual stock, a plaque in the actual resin) and write the tolerance as a ΔE against that standard. The same Pantone reference will land at different L*a*b* values on a coated carton, a polypropylene label and a moulded cap.
Why does my factory’s Delta E differ from my own reading?
Almost always because one of the fixed conditions is not the same: formula, illuminant and observer, instrument geometry or SCI/SCE setting, aperture, sample preparation or the standard itself. Two spectrophotometers from different makers can also disagree by several tenths of a ΔE on the same sample. Agree one preparation, exchange a set of ceramic reference tiles to check inter-instrument agreement, and compare numbers only after both laboratories measure the same sealed standard.
Should a colour tolerance be in the contract or in the specification?
Both, by reference. The quality specification or technical schedule carries the detail (standard, formula, limit, method, sampling, reporting); the supply agreement states that batches must conform to that specification and sets out the consequence of non-conformance and the dispute route. Putting the numbers only in an email thread leaves them outside the agreement; putting them only in the contract makes every method change a contract amendment.
Does a Delta E limit cover colour change during shelf life?
Only if the stability protocol includes one. A release tolerance compares a fresh batch with the standard; a shelf-life tolerance compares the product with itself over time, usually at the stability pull points. Set both for products that are known to yellow or brown, including white emulsions and formulas with vitamin C, retinoids, natural extracts or certain fragrances, and record L*a*b* at each pull rather than a subjective note of “slight change”.
Sources and limitations
Primary and technical sources used in this article, all checked on 30 October 2026: the CIE publication pages for ISO/CIE 11664-1 (standard colorimetric observers), 11664-4 (CIE 1976 L*a*b* colour space) and 11664-6 (CIEDE2000 colour-difference formula), and the BSI product page for ISO/CIE 11664-2 (standard illuminants); ASTM D2244-23 (calculation of colour tolerances and colour differences), ASTM D1729-22 (visual appraisal of colours and colour differences) and ASTM E1347 (colour measurement by tristimulus colorimetry), as described on ASTM’s standard pages; the Konica Minolta Precise Color Communication guide and the X-Rite tolerancing series as instrument-maker explanations of colour spaces and tolerancing; the Pantone colour-systems page; 21 CFR Parts 73 and 74 Subpart C on eCFR and the FDA colour additive summary table; the European Commission’s cosmetics legislation page and CosIng database; the UK government’s guidance on making cosmetic products available in Great Britain; and the Health Sciences Authority of Singapore’s ASEAN Cosmetic Directive page. The CIEDE2000 values in the worked examples were calculated by us using the published procedure and verified against the CIE reference test pairs; ISO 12647-2 and ISO 2859-1 are referred to by name because their full texts are paid standards.
Limitations: every tolerance, L*a*b* value and ΔE in this article is illustrative and chosen to show the arithmetic; none is a measurement supplied by a named manufacturer or a requirement of any standard. The perceptibility bands are a widely used convention rather than a published threshold, and real visibility varies with colour, observer, lighting and context. Instrument behaviour, print standards and colourant regulations change; confirm the current edition of any standard with its publisher and the permitted-colourant status of any pigment with the regulator of each market you sell into. Colour tolerances must finally be set from your own factory’s measured batch history, in writing, before the first production order.
This article is independent editorial content provided for general information. It does not constitute legal, regulatory or technical advice. Verify all figures with your manufacturer and a qualified adviser before making commercial decisions.