Delta E is the distance between two colours in a colour space, expressed as a single number. It exists because “close enough” is not something two companies can agree on across a telephone, and because the human eye is simultaneously extraordinarily sensitive and completely unreliable as a record.
If you buy or sell coloured plastic, this is the number your commercial relationship should be built on. Here is what it actually means and where it misleads.
The colour space underneath
Delta E is measured in CIELAB, which places any colour at a point defined by three coordinates:
– L\ – lightness, from 0 (black) to 100 (white) – a\ – the green-to-red axis, negative green, positive red – b\* – the blue-to-yellow axis, negative blue, positive yellow
CIELAB was designed to be perceptually uniform: the intention is that equal distances in the space correspond to equal perceived differences. It gets close, and where it falls short is exactly why there is more than one Delta E formula.
The formulas, and why the version matters
dE76 (CIE 1976) is straight Euclidean distance:
dE76 = the square root of (ΔL\)² + (Δa\)² + (Δb\*)²
Simple, still widely used, and it overstates differences in saturated colours, particularly blues. Two deep blues that look identical can return a dE76 of 3.
dE94 adds weighting factors for chroma and hue, correcting the worst of the saturation problem. It has separate parameter sets for textiles and for graphic arts.
dE2000 (CIEDE2000) is the current recommendation and the most perceptually accurate. It adds corrections for lightness, chroma and hue dependence, plus an interaction term for the blue region. It is more complicated to compute, which no longer matters since the instrument does it.
These numbers are not interchangeable. The same pair of samples can give dE76 of 1.8 and dE2000 of 0.9. A specification that says “dE less than 1.0” without naming the formula is not a specification – it is the beginning of an argument.
Write the formula into the contract. “ΔE\*00 ≤ 1.0, D65/10°, specular included” is a specification. “Delta E under 1” is a wish.
What tolerance is realistic
There is no universal answer, because acceptability depends on where the part sits and what it sits next to. But there are useful reference points.
| Delta E (dE2000) | What it means in practice |
| Under 0.5 | Not perceptible to almost anyone, side by side |
| 0.5 to 1.0 | Perceptible to a trained observer under controlled light |
| 1.0 to 2.0 | Perceptible to a careful observer; usually acceptable for most industrial parts |
| 2.0 to 3.5 | Clearly visible side by side; acceptable only where parts are never adjacent |
| Over 3.5 | Obviously different colours |
The context that decides your number:
Parts that touch each other – a housing and its adjacent bezel, a cap on its bottle – need tight tolerance, because the eye is far better at detecting a difference across a joint than in isolation.
Parts that are never seen together – crates in a warehouse, pipe underground – can tolerate a great deal more.
Large flat surfaces show variation more than small textured ones.
Light and pastel shades show a given Delta E more visibly than deep saturated ones – and simultaneously are harder to hold, because a small pigment variation moves them further.
Being honest about achievability matters more than being ambitious. A tolerance tighter than the process can reliably deliver produces a stream of technically-failing, commercially-acceptable lots and a supplier relationship built on waivers. Agree a number both sides can live with and then hold it.
Measurement conditions, which are half the specification
Two laboratories can measure the same plaque and disagree, entirely legitimately, because they measured differently. Fix these in writing:
Illuminant. D65 – average daylight – is standard for most industrial work. Others exist for specific retail and automotive contexts. The illuminant matters enormously because of metamerism, below.
Observer angle. 10° is standard for larger fields of view; 2° for small ones. They give different numbers.
Specular component. Specular Included (SCI) measures total reflectance and largely ignores surface finish, which makes it the right choice for judging the colourant. Specular Excluded (SCE) removes the gloss reflection and correlates better with how the part looks to a person. Neither is wrong; they answer different questions, and comparing an SCI reading to an SCE reading is meaningless.
Sample preparation. This is the one most often left undefined and it moves results more than the instrument does. Injection-moulded plaque or compression-moulded? What thickness? What surface finish? Measured on the gate side or the far side? Conditioned how long after moulding? A colour measured on a 2 mm plaque is not the same colour on a 4 mm plaque, because opacity changes with path length.
Number of readings. Multiple points, averaged. A single reading on a moulded plaque is a sample of one on a surface that is not perfectly uniform.
Metamerism: the trap
Two samples can match perfectly under one light source and differ obviously under another. This is metamerism, and it happens when the two colours are built from different pigments whose spectral reflectance curves cross rather than coincide.
It is the reason a part approved in the lab under D65 is rejected on the shop floor under fluorescent tubes, or in the customer’s showroom under LED.
A low Delta E under a single illuminant does not prove a match. Where the application matters, specify measurement under at least two illuminants – typically D65 and an incandescent or store-lighting source – and set a limit on the metamerism index.
The fundamental defence is a spectral match rather than a colorimetric one: matching the reflectance curve itself, which usually means using the same pigment chemistry as the standard rather than a clever combination that lands on the same coordinates. That is a formulation decision, and it is one to raise when the standard is set, not after a rejection.
What to put in a colour specification
– The standard: a physical plaque, a spectral data file, or both. A physical standard should have a stated issue date and a replacement interval, because standards fade. – Formula and tolerance: e.g. ΔE\00 ≤ 1.0 – Illuminant and observer: e.g. D65/10° – Specular: included or excluded – Sample preparation: process, thickness, finish, measurement location, conditioning time – Metamerism requirement if the part is viewed under more than one light source – Individual coordinate limits where a directional shift matters – ΔL\ alone, for instance, if going darker is acceptable but going lighter is not – What happens on a marginal lot: who decides, and how
That last point prevents most disputes. Agreeing the arbitration route in advance costs nothing; agreeing it during a rejection costs a great deal.
We set out the wider version of this in how to write a masterbatch specification, and the causes of drift in why colour drifts between batches.
Things that move colour that are not the masterbatch
Before a colour complaint becomes a supplier complaint, rule these out:
– Base resin change – a different grade or supplier has a different natural colour – Regrind percentage – regrind is thermally aged and darker or yellower – Filler loading – calcium carbonate is a white pigment in its own right – Melt temperature or residence time – some pigments shift with thermal history – Part thickness – opacity changes, so shade changes – Surface finish – a change in mould texture changes measured and perceived colour – Dosing accuracy – a volumetric feeder drifting by a few percent moves shade
Every one of these has ended with a masterbatch being blamed for a change made somewhere else.
Matching in-house with single-pigment concentrates
Some processors prefer to hold colour control themselves rather than buy a finished shade. Mono concentrates – single-pigment concentrates dosed in combination – make that possible: you keep a small set of primaries, adjust ratios against your own measurement, and respond to a customer shade change without a new formulation lead time.
The trade-off is real. You take on the colour matching responsibility, you need the instrumentation and the discipline to use it, and metamerism becomes yours to manage. For a converter with a spectrophotometer, trained staff and frequent short-run colour changes, it is often the better model. For a plant running three colours a year, buying a matched colour masterbatch is simpler and cheaper.
Talk to us
We match to a physical standard or to spectral data, under agreed conditions, and manufacture under ISO 9001:2015.
Send us your standard, your polymer and grade, your part thickness and the illuminants the part will be seen under, and we will tell you what tolerance is realistic before you commit it to a purchase order. Related: white masterbatches, special effect masterbatches, cosmetics, automotive.
FAQs
What is Delta E in colour measurement? A single number expressing the distance between two colours in the CIELAB colour space, where L* is lightness, a* the green-red axis and b* the blue-yellow axis. It converts a visual judgement into a measurable quantity.
What is an acceptable Delta E for plastics? It depends on the application. Under 1.0 is typical where parts sit next to each other, 1.0 to 2.0 is acceptable for most industrial parts, and 2.0 to 3.5 can be fine where parts are never seen together. Set the number against how the part is actually viewed.
What is the difference between dE76, dE94 and dE2000? dE76 is plain Euclidean distance and exaggerates differences in saturated colours, especially blues. dE94 adds chroma and hue weighting. dE2000 adds further corrections and is the most perceptually accurate. The same samples give different values under each, so the formula must be stated.
What does D65/10 degree mean in a colour specification? D65 is the standard daylight illuminant used for the measurement, and 10 degrees is the standard observer field of view. Both change the result, so both belong in the specification alongside the tolerance.
What is metamerism? When two samples match under one light source and visibly differ under another, because they are made from different pigments with different spectral reflectance curves. It is why a lab approval under daylight can fail under shop-floor fluorescent lighting.
What is the difference between specular included and specular excluded? Specular included measures total reflectance and largely ignores surface finish, making it the better measure of the colourant itself. Specular excluded removes the gloss reflection and correlates better with visual appearance. They are not comparable to each other.
Why does the same masterbatch give a different colour in my part? Usually something other than the masterbatch changed – base resin grade, regrind percentage, filler loading, melt temperature, part thickness or surface texture. Each of these moves measured colour, and all are common causes of a complaint that lands on the colourant.
Should I buy matched colour masterbatch or mono concentrates? Buy matched masterbatch if you run few colours and want the supplier to carry the matching responsibility. Use mono concentrates if you run frequent short colour changes, have a spectrophotometer and trained staff, and want to control shade in-house without waiting for a new formulation.