Colour Management Across Screen and Print, and Why Matching Fails

The oldest complaint in the trade

The proof does not match the screen. The printed brochure is duller than the file. The brand red on the website is not the brand red on the packaging. These complaints arrive constantly, and they are usually met with the assumption that something is broken. In most cases nothing is broken. Two devices producing colour by opposite physical means, within different ranges, are being asked to agree, and the surprise should be that they ever come close.

Colour management is the discipline built to make them come as close as physics permits, and to make the residual differences predictable rather than random. Understanding what it can and cannot do converts an intractable argument into a set of decisions.

Two incompatible ways of making colour

A display is additive. It emits light from red, green and blue elements, and mixes by summing. All three at full output give white; all off give black, or the nearest the panel manages. The device generates its own light, and the room contributes nothing to the colour except by adding flare.

Print is subtractive. Ink removes wavelengths from light that falls on the paper and is reflected back. Cyan, magenta and yellow inks each subtract a portion; combined they should approach black, though in practice they yield a muddy brown, which is why a separate black ink, the K in CMYK, is added for density and for neutral shadow control. Nothing is emitted. The maximum brightness available is the white of the substrate, and the maximum darkness is limited by how much ink the surface will accept.

The consequences are immediate. A screen with a backlight can produce a luminous saturated blue that no combination of inks on paper can reach, because ink cannot emit. Paper can produce a soft, warm neutral in a tinted uncoated stock that a display renders as slightly grey and lifeless. These are not calibration errors. They are the boundaries of two different physical systems.

Gamut, and the impossibility of complete agreement

The gamut of a device is the set of colours it can actually produce. Plotted in a colour space, each device occupies an irregular volume, and no two are identical.

  • sRGB. The default working space of the web and most consumer displays. Modest in extent, particularly in cyan and green, but universal, which is its real value.
  • Adobe RGB (1998). Wider in the greens and cyans, historically chosen because it encompasses more of what CMYK can print in those regions.
  • Display P3. Wider in reds and greens than sRGB, now standard across much modern hardware, and the reason images increasingly look more vivid on a phone than on an older monitor.
  • CMYK process print. Not a single gamut but a family, differing by ink set, press, screening and above all substrate. Coated stock holds ink on the surface and yields denser, more saturated colour; uncoated stock absorbs it and lightens everything.

Crucially these volumes overlap without containing one another. A print gamut is smaller overall but exceeds sRGB in certain yellows and oranges. Any workflow that assumes print is simply a reduced version of screen will be wrong in exactly those regions.

Profiles and the profile connection space

The framework that manages this is the ICC system, and its central idea is indirection. Rather than defining a conversion from every device to every other device, which would require an unmanageable number of mappings, each device is characterised once against a shared reference.

An ICC profile is a file describing how a specific device’s numbers relate to actual colour. A display profile says what colour appears when the panel is fed a given RGB triple. An output profile says what colour results when a press lays down a given CMYK combination on a given paper. Both are expressed relative to the profile connection space, a device independent space based on CIE XYZ or CIELAB.

Conversion therefore happens in two stages. Source numbers are translated into the connection space using the source profile, then translated out into destination numbers using the destination profile. The colour management module performs the arithmetic. Because everything meets in the middle, adding a device requires one new profile rather than a new mapping for every existing device.

Where profiles come from

A display profile is made by calibration followed by characterisation. Calibration sets the display to a defined state: white point, luminance, tone response. Characterisation then measures a series of patches with a colorimeter or spectrophotometer and records the relationship in the profile. The first step changes the device; the second describes it. Both are required, and profiling an uncalibrated display simply documents an arbitrary condition that will drift.

An output profile is made by printing a test chart of hundreds or thousands of patches under production conditions, measuring each, and building the transform from the results. It is valid only for those conditions. Change the paper, the ink or the press settings and the profile no longer describes reality, which is why generic profiles supplied with software are a starting point rather than an answer.

Rendering intents: what to do with the unreachable

When a source colour falls outside the destination gamut, something must give. The rendering intent selects the strategy.

  • Perceptual. Compresses the whole gamut smoothly so that relationships between colours are preserved. Every colour shifts slightly, including reproducible ones, but gradations survive without banding or flattening. The usual choice for photographic images with significant out of gamut content.
  • Relative colorimetric. Reproduces in gamut colours exactly and clips out of gamut colours to the nearest reproducible equivalent. White is mapped to the substrate white. Accurate where it can be, and prone to losing detail in saturated areas where several distinct colours collapse to the same clipped value.
  • Absolute colorimetric. Like relative, but reproduces the source white point rather than adapting to the destination. Used for proofing, where the aim is to simulate another substrate, including its paper tint.
  • Saturation. Prioritises vividness over accuracy. Appropriate for charts and diagrams where impact matters more than fidelity, and inappropriate for anything photographic.

No intent is correct in general. Perceptual applied to a corporate logo will shift a brand colour that was perfectly reproducible; relative colorimetric applied to a sunset will flatten the sky into a band of identical orange.

Why matching fails in practice

Assume competent profiles throughout, and matches still fail. The reasons are mostly mundane.

  • Untagged files. An image without an embedded profile carries numbers with no stated meaning. Software assumes a space, usually sRGB, and if the file was authored elsewhere the assumption is wrong from the first operation.
  • Uncontrolled viewing conditions. A proof judged under a warm office lamp, then under a window, then in a D50 booth will look like three different proofs. Print has no light of its own, so its appearance is at the mercy of the room. This is the single most common cause of disputes.
  • Adaptation. The visual system continuously renormalises to its surroundings. A screen at 250 candelas per square metre next to a print lit at a fraction of that intensity will never be judged fairly, because the eye cannot be adapted to both simultaneously.
  • Substrate difference. A profile describes a paper. Print on a different stock and the white point, the ink holdout and the maximum density all change, and the profile is describing a press run that is not the one happening.
  • Optical brighteners in paper. Many stocks contain fluorescent agents that absorb ultraviolet and emit blue. Under a UV rich source the sheet looks bluer than under a booth with UV filtered out, and the measurement disagrees with the eye.
  • Spot colours forced through process. A specified spot ink converted to CMYK will land wherever the process gamut permits. Many of the most common brand colours, particularly saturated oranges, deep blues and vivid greens, are outside it.

What a working setup looks like

The remedies are procedural rather than clever, and they are the difference between a workflow that argues and one that ships.

  • Calibrate displays on a schedule. Backlights drift. Monthly is reasonable; before any critical approval is better. A profile built once and trusted for two years is worse than none, because it invites false confidence.
  • Standardise the viewing light. A D50 booth for print assessment, positioned so the display is not competing with a window. Control the surround: a bright red wall behind a monitor will bias every judgement made in front of it.
  • Embed profiles everywhere. Every file, every stage. The cost is a few kilobytes and the benefit is that numbers retain their meaning across handoffs.
  • Soft proof, then hard proof. Soft proofing simulates the output on screen with gamut warnings, which catches problems early. It does not replace a contract proof produced on a calibrated printer with the correct profile and assessed under standard light.
  • Design within the destination. The most reliable fix is chosen before production. Brand palettes verified as reproducible in the intended output, rather than picked on a wide gamut display and defended afterwards, remove the problem instead of managing it.
  • Set tolerances, in numbers. A ΔE limit on key colours, measured on the delivered sheet, turns approval into a measurement rather than an opinion held across a table.

Colour management does not deliver identity between screen and print. It delivers predictability: a known relationship, a defined set of compromises, and a difference that was decided deliberately rather than discovered at delivery.