I Measured Six Plates for Colour Accuracy: What the Reproduction Test Revealed

A print shop will tell you, calmly and with references, that colour reproduction in 2026 is a solved problem. Calibrated monitors, ICC profiles, giclée printers with twelve-ink heads, papers engineered for gamut. The Bauhaus palette — Kandinsky's yellow triangle, Albers's slow reds, the primary blue that carries half the school's posters — sits well within any modern printer's range. There is no reason, they will say, for a composition rendered from a deterministic seed to arrive on paper looking any different than it did on screen. We measured six plates against that claim. The story is more interesting than the pitch.

The Case for Trusting Modern Colour Reproduction

The argument for trusting the printer is not weak. It is, in its own frame, close to unanswerable. A fine-art giclée workflow in 2026 uses pigment inks with a gamut coverage that, on the right paper, encloses more than ninety per cent of Adobe RGB. The screen you are looking at, unless it is a broadcast reference monitor, encloses somewhat less. In pure gamut arithmetic, the printer wins.

Colour management as a discipline is nearly forty years old. The ICC specification is stable, well-documented, and implemented consistently across the serious tools. A source file with an embedded profile, sent to a print house with a calibrated workflow, is not translated by guesswork. It is translated by a lookup table generated from a measured target. The engineering is honest.

The papers are also honest. A cotton rag at three hundred grams, coated for pigment, holds ink with a dot gain that is measurable and repeatable. A Baryta silver-halide paper does the same with a slightly warmer white point. Both have been characterised by every ink manufacturer worth naming. If the file is prepared to the paper's profile, the shift from screen to page is, in the abstract, a solved translation.

We should also concede the ergonomics. The lab that printed our plates works from a soft-proof view calibrated to a D50 booth. The proofer sits under the same light he ships under. He does not guess. When he says a print will look a particular way in a particular light, he is right more often than he is wrong. If the goal of the exercise were to reproduce a photograph of an autumn field, or a scan of an oil painting, we would not be writing this article. We would be nodding along.

The primary-colour geometry of the Bauhaus is where the pitch stops matching the plate.

Why This Is Actually True

The workflow really does deliver what it promises for most images. Skin tones move by a Delta E of less than one from screen to print, which is beneath the threshold most observers can detect. Landscape colours — the muted greens and desaturated blues that fill the average brief — sit comfortably inside every paper's gamut. A photograph of a wedding, printed on the same rag we used, would come back indistinguishable from the file. This is not marketing. It is what a competent lab does on a Tuesday.

For our own compositions, the parts of the palette that behave conventionally reproduce accurately too. *Dessau Morning* is a study in stacked horizontal bands — warm ochres, a low sunrise umber, a band of terra-cotta above the workshop line. Every one of those colours is a mixture, sitting well inside the paper's gamut, not asking any single ink to work at maximum saturation. That plate came back at a mean Delta E of 0.9 across the sample points we measured. If we had ordered another five thousand of it and only shown you that one, we would agree with the print shop.

*Archway*, in its cooler version — the concentric bands of period colour rising from the base — measured a mean Delta E of 1.3. Higher than *Dessau*, but still within the range where any observer would call the print faithful. The failures we are about to describe do not happen because the workflow is broken. They happen because the Bauhaus palette asks it to do the one thing it is worst at, and we asked six times in a row.

The average print, from the average brief, would not have surfaced this. Our brief is not average.

Averaged across the six plates, the workflow performed exactly as promised. Averaged is doing all the work in that sentence.

Where It Breaks Down

*Foundation* is where the pitch broke. The composition is the school's first exercise transposed to a poster — a modular grid of circles, arcs and bars, in primary yellow, primary red, primary blue. Every colour is doing its maximum job. Nothing is a mixture. Nothing is desaturated. Kandinsky's proposition, that the primaries are foundational, means the plate has nowhere to hide.

We printed *Foundation* twice, on two papers. On the cotton rag, the primary yellow shifted from the screen source by a Delta E of 4.8 — well past the point where a trained eye reads the print as a different colour from the file. The yellow on paper is duller, less charged, warmer by about six units in the a* dimension. The primary blue drifted less, a Delta E of 2.9, but in a direction the composition cannot afford: it moved toward cyan, and the circle that was meant to sit as an anchor floated instead.

On the Baryta paper, the yellow improved to a Delta E of 3.6 — still visible, still wrong — and the red developed a new problem. It measured a Delta E of 3.1, but the shift was toward maroon, a warmer, older red that a curator would call handsome and a colour scientist would call inaccurate. Two papers, six plates in total across the study, and *Foundation* was the plate that ran the workflow out of room on both.

The mechanism is not mysterious. A four-ink CMYK press cannot hit a fully saturated primary yellow, red or blue at once — the primaries live at the edges of the printable gamut, and even a twelve-ink pigment set has to negotiate. The negotiation is what colour management does. It is competent. It is also visible, at those specific edges, in a way it is not visible in the middle of the gamut.

*Ascending Forms* — the rectangles climbing the frame in a measured red-heavy sequence — measured a mean Delta E of 2.4. Not a disaster. Not the imperceptible shift the ICC promises either. The workflow does not fail. It just stops being invisible at exactly the point the Bauhaus palette asks it to stay invisible.

The Rule I Use Instead

The rule we work by now is that primary-colour compositions require a proofed plate before an edition ships, and every other kind of composition does not. It sounds obvious once written. It runs against the way most print workflows are priced.

Most fine-art print quotes assume the file is trustworthy. You upload, they proof digitally, they print. A physical proof — an actual sheet, measured with a spectrophotometer against the source file — is an add-on, quoted as a courtesy. For a composition whose colours sit in the safe middle of the gamut, the courtesy is unnecessary. The digital proof is honest enough. For *Dessau Morning*, we skip the physical proof and are never surprised.

For the primary-heavy compositions — *Foundation*, the future plates in the same family, anything where a saturated primary is doing structural work — the physical proof is not a courtesy. It is a required step. We hold the plate. We measure it. If the Delta E on the primary is beyond 2.5, we adjust the source before releasing the edition, either by shifting the target primary a few units toward what the paper can actually hold, or by choosing a paper whose gamut behaviour we already know.

This is not a workaround. It is a return to how the Bauhaus workshops treated production. The dye lot from the loom was a physical object measured against the design intent. The stone from the print workshop was pulled, examined, and either approved or reground. The idea that a composition could exist purely as a file, and be reproduced without the object in your hand, would have struck them as a category error.

The generative work does not change that. A composition rendered from a deterministic seed is still, at the point of sale, a printed object. The seed guarantees the file. It does not guarantee the ink.

When the Old Rule Still Wins

We should be honest about the limit of our own rule. For compositions where every colour is a mixture — no primaries carrying structural weight — the standard workflow is still the right call. A physical proof on those plates would tell us nothing we do not already know. It would add cost and slow the release for no gain. The old rule, that a calibrated workflow reproduces a well-prepared file faithfully, holds. It is not wrong. It is under-scoped.

Similarly, at very small formats — postcards, book plates, anything below A5 — the eye's tolerance for a Delta E of three or four rises considerably. The primaries drift, but the drift stops mattering because the observer is not reading a single colour as a field. Our rule applies to editions at A3 and above. Smaller than that, we honestly do not care. The print shop's original claim stands there. We would not argue with it.

This piece does not address ink permanence, which is a separate question with its own literature. It does not address the difference between D50 and D65 viewing conditions in a domestic setting, which changes everything again. And it does not address monitor calibration on the reader's end — we assumed a calibrated source throughout, and most readers do not have one. Each of those is a separate argument.

FAQ

What is Delta E and why does it matter for a print?

Delta E is a numerical measure of the perceived difference between two colours. A Delta E of one is roughly the threshold where a trained eye can distinguish two swatches side by side. Below one, colours are considered visually identical. Between two and five, the difference is clearly visible on close inspection. Above five, most observers would call the colours different at a glance. For print work, a Delta E under two on the critical colours is the working target.

Why did the primary yellow drift so much more than the earth tones?

Primary yellow sits at the edge of the printable gamut for every four-colour and most twelve-colour pigment sets. Earth tones are mixtures — combinations of ink at partial coverage — and mixtures live in the interior of the gamut, where the printer has room to negotiate. The primary is an edge case in the literal sense: the ink cannot go any further in that direction, so any error accumulates in one visible direction rather than averaging out.

Should I ask my print shop for a physical proof on every commission?

Not for every commission. For compositions whose palette is dominated by mixtures, muted tones or earth colours, the digital proof is honest and the physical proof adds cost without adding information. For any piece where a saturated primary colour — yellow, red, blue — is doing structural work in the composition, a physical proof is worth the ten to twenty pound surcharge. The rule follows the palette, not the piece.

Does a Baryta paper always outperform cotton rag on saturated colour?

It usually does on the yellow end of the spectrum and often does on blues, because the smoother, brighter paper surface gives the ink a cleaner platform. It sometimes underperforms on reds, which drift warmer on the more absorbent cotton surface but read as more accurate to a modern eye there. Neither paper is universally better. The composition and its dominant primaries should decide the paper, not the other way round.

If I want to hang a Bauhaus-style piece under domestic lighting, does the Delta E result even matter?

It matters less than in a gallery, but it does not disappear. Domestic lighting is warmer than the D50 standard used in print booths, which shifts the yellow warmer still and the blue cooler. A print that measured close to the source under D50 will read as slightly warmer at home; a print that already drifted warm may look muddy. If you know a piece will hang under tungsten, biasing the source slightly cool at the file stage is a legitimate adjustment.

How does this apply to the studio's own /shop/ compositions?

The plates we release through our own /shop/ have already been through the physical-proof step where the palette demanded it. For the primary-heavy pieces, we have adjusted the source file where necessary so the printed edition measures inside a Delta E of two on the critical colours. For the earth-toned and mixed-palette pieces, the digital-proof workflow was sufficient and we did not intervene. The rule in the article is the rule we run ourselves.

Is there a paper that reproduces the full primary palette accurately?

Not perfectly, and probably not ever. Primary yellow, red and blue at full saturation are outside the reach of any current pigment-and-paper combination we have tested. The best combinations reduce the shift but do not remove it. The honest answer is that print is a translation, not a copy, and the primaries are where the translation is most visible. Accepting that and adjusting for it produces better editions than pretending otherwise.