Lab, Delta E, and what “close enough” means

A number that means the same thing everywhere

R 200 G 30 B 40 is an instruction to a device, not a description of a colour. It means “drive your red channel hard”, and what comes out depends on the device.

To talk about colour independently of any device, you need a space defined by measurement rather than by hardware. CIELAB is the one everyone uses:

• L\*: lightness, 0 (black) to 100 (white)
• a\*: green to red
• b\*: blue to yellow

Lab is defined from measurements of how human vision actually responds, and it is perceptually roughly uniform: a given distance in Lab corresponds to roughly the same perceived difference anywhere in the space. That uniformity is what makes the next idea possible.

Delta E

ΔE is the distance between two colours in Lab. It is the number the whole print industry argues in.

Rough interpretation, for the commonest formula (ΔE*ab, or CIE76):

• Under 1: not distinguishable by a human eye, even side by side.
• 1 to 2: visible to a trained eye in a side-by-side comparison.
• 2 to 3.5: visible to anyone in a side-by-side, usually invisible when seen separately.
• 3.5 to 5: obviously different side by side.
• Over 5: a different colour.

Later formulas (ΔE94, ΔE2000) correct known weaknesses of CIE76, particularly that it overstates differences in saturated blues. ΔE2000 is the current standard and gives smaller numbers than CIE76 for the same pair, so a supplier quoting ΔE must say which formula. Ask.

What tolerance to actually ask for

The international standard for offset printing, ISO 12647-2, sets tolerances for the solid primaries. In broad terms it permits a deviation of around ΔE 5 for the process colours on a production run, with tighter figures for the OK sheet against the proof.

For practical conversations:

• A logo colour that matters: ask for ΔE ≤ 2 against the supplied reference, and expect to pay attention to it on press.
• General four-colour work: ΔE ≤ 5 is normal and nobody will notice.
• Across two different suppliers: do not promise better than ΔE 3 without spot inks.

Saying “ΔE 2 against Pantone 2925 C measured under D50” is a specification. Saying “it must match exactly” is not, and the shop cannot act on it.

Measuring it, free and not free

• A spectrophotometer: an X-Rite i1 or similar, several hundred pounds new and much less used. It measures a printed patch and gives you Lab. Shops have one; ask them to measure your logo on the sheet.
• Free calculation, once you have two Lab values: any ΔE calculator, or a few lines of Python. colormath and colour-science both implement ΔE2000.
• Free and rough: photograph a printed sample and your reference side by side under the same light, open in GIMP, and use the colour picker. This tells you whether there is a difference, not how much, because your camera and your screen are both unprofiled. It is still better than arguing from memory.

Where you will actually use this

You will not compute ΔE often. You will use the vocabulary constantly:

• To ask a supplier what tolerance they work to, and to hear a real answer.
• To decide whether a reprint is justified, rather than arguing about whether the blue “looks right”.
• To specify a brand colour in a usage sheet as Lab as well as CMYK, RGB and hex, because Lab is the only one of the four that means the same thing in two different factories.

That last is the concrete takeaway. Put a Lab value in your brand sheet. Almost nobody does, and it is the value a printer in another city can actually hit.

One thing to be careful about

A ΔE figure quoted without a measurement condition is incomplete. Two labs measuring the same printed patch can report different Lab values depending on the instrument’s illuminant, its filter, the backing behind the sheet, and whether the paper contains optical brightening agents that fluoresce under UV. The industry standard for this is M1, which specifies a defined UV content in the measuring light, and it exists precisely because brightened papers gave inconsistent readings under the older M0 condition.

You do not need to own an instrument to use this. You need to know that “ΔE 1.8” from one supplier and “ΔE 1.8” from another are only comparable if both were measured the same way, which is a fair question to ask, and a supplier who can answer it is telling you something useful about how they work.