The first run was signed off in March and looked right. The repeat lands in September, the brown is warmer, the corporate blue has gone slightly green, and your retail buyer photographs the two boxes side by side and asks which one is correct. Carton print colour drift is not usually a failure of care — it is what happens when a colour was approved by eye against an artefact nobody defined, on a substrate that changes batch to batch, printed by a process with a dozen live variables and no agreed number to measure against.
The fix is not "be more careful". It is to replace the approval ritual with a small set of recorded, measurable things, and to write a tolerance that names what is being measured, on what, under which conditions.
Why corrugated drifts more than you expect
Direct flexo post-print on corrugated board is a coarse process compared with the offset work your artwork was probably designed in. Six variables move between March and September, and most of them are invisible on a job ticket unless someone writes them down.
The liner itself. Recycled testliner shade tracks the recovered-fibre furnish going into the mill. Two reels bought six months apart can differ visibly in brown before a drop of ink touches them, and every ink laid over that liner shifts with it. Board is a substrate with its own colour, not a white sheet.
Board caliper and washboarding. Flutes telegraph through the liner. Where the liner dips between flute tips, ink lays down differently, so the same solid reads as a striped solid under raking light — and a spectrophotometer aimed at the wrong spot reads a number that has nothing to do with the ink.
Anilox condition. The anilox roll meters ink by cell volume, quoted in BCM (billion cubic microns per square inch) and line screen. Cells wear and score, volume falls, solids go weak. A repeat run on a different anilox from the first is a different colour, no matter how carefully the ink was matched in the bucket.
Plate and impression. Photopolymer plates wear, and impression pressure set too heavy dot-gains the mid-tones and crushes the flute. A press operator chasing a solid by adding impression is trading colour accuracy for board strength.
Ink chemistry on press. Water-based flexo inks run at a controlled viscosity, checked with an efflux cup and quoted in seconds, and at an alkaline pH held by amine that evaporates over a shift. As pH falls the resin solubility changes and the colour shifts — a well-run press checks and corrects it hourly.
Speed and drying. The same ink at a different press speed transfers differently. A run that started at one speed and finished faster to make a sailing date does not print the same colour end to end.
None of this is exotic. It is why the useful question to a supplier is not "can you match this?" but "what will you record so the next run repeats this one?"
The artefact you are matching to
Most colour disputes are really disputes about the reference. Four things get called "the sample", and they prove different amounts.
| Reference | What it is | What it proves | What it does not prove |
|---|---|---|---|
| Digital / contract proof | Calibrated inkjet or press proof on proofing stock | The artwork's intended appearance and separations | Nothing about your liner, ink set, anilox or press |
| Drawdown | Ink hand-drawn on the actual liner with a calibrated bar | The ink formula's colour on your real substrate | Nothing about press transfer, dot gain or solid density at speed |
| Press OK / first-off sheet | A sheet pulled and signed during the actual run, at running speed | The whole system reproduced the target once | That the rest of the run held it |
| Retained master | A signed, dated production sheet held by both parties, with measured values | The agreed reference for every future run | Nothing, once it has faded — see below |
The retained master is the one that matters, and it is the one most often handled badly. Three details make it work:
- Three copies, not one. One at the factory, one with you, one sealed and never opened except to settle a dispute.
- Store it dark and sealed. Paper yellows and inks fade under UV. A master pinned to a wall for two years drifts on its own, and every subsequent run is matched to a target that has moved. Opaque bag, sealed, in a drawer.
- Record its numbers. A master with measured L*a*b* values for each colour-critical patch survives being lost, burned or aged. A master without them is a piece of card with an opinion attached.
Re-issue the master on a stated interval — annually is a common cadence — by pulling and signing a fresh sheet from a run that measures inside tolerance against the recorded values, not against the old physical sheet.
Writing a tolerance that means something
"ΔE ≤ 2" written on a PO is close to meaningless. Six qualifiers turn it into something a laboratory could arbitrate.
Which ΔE. ΔE*ab (CIE 1976) and ΔE00 (CIEDE2000) give different numbers for the same pair of samples, and CIEDE2000 is generally the one that tracks what an eye sees. Name the formula.
Illuminant and observer. D50 with the 2° standard observer is the graphic-arts convention. If you specify D65 and your supplier measures D50, you will disagree about a colour you both measured correctly.
Measurement condition. ISO 13655 defines M0, M1, M2 and M3. The difference is UV content in the instrument's illumination, and it matters enormously on white-top liners containing optical brightening agents. Measure the same OBA-loaded sheet at M0 and M1 and the numbers diverge by more than most tolerances allow. This single line is the most common cause of two parties measuring the same box and getting different answers. Specify one condition — M1 is the modern default — and require both sides to use it.
Backing. Black backing or white backing behind the sample changes the reading on lightweight or translucent stock. Name it.
Where on the sheet. On fluted board, define measurement on a solid patch positioned off the flute tips, or on a designated colour bar, and require an average of a stated number of readings rather than a single spot. Measuring across washboarding produces noise that gets blamed on the ink.
Which elements. Not every element on a box is colour-critical. A brand solid and a regulated warning panel are; a background rule is not. Mark colour-critical elements in the artwork file itself and give them the tight tolerance, with a looser one everywhere else. One tolerance applied to the whole box is either unenforceable or absurdly expensive.
Then state three separate numbers, because they are three separate problems: sheet-to-sheet variation inside one run, first-off to last-off across one run, and run-to-run against the retained master. Run-to-run is the one that produced the September box.
The brown-liner problem
Pantone references are defined on white coated and uncoated stock. Printed on kraft or recycled brown testliner, a PMS colour cannot be hit — the substrate is subtractive and the ink is not opaque. There are only three honest routes: accept a shifted target and record its measured L*a*b* on your actual liner as the reference; specify a white-top liner; or add a white flood underprint, which costs a print unit and a pass. Any supplier promising a clean PMS match on brown board without one of those three is describing an outcome they cannot control. Ask which route is being proposed before the quote is compared to anyone else's — it changes the price.
The recipe card
The single highest-value thing you can ask for is that the press records the conditions of the approved run and re-uses them. Call it a recipe card, a job ticket, a control sheet — the name is irrelevant, the fields are not.
| Field | Why it decides the repeat |
|---|---|
| Liner grade, grammage and supplier or mill batch | Substrate colour is the base every ink sits on |
| Flute profile and board combination | Washboarding and ink lay change with flute |
| Ink formula or ink-supplier reference number per colour | "Same blue" is not a formula |
| Anilox line screen and BCM per unit | The biggest single press-side driver of solid density |
| Plate set ID, durometer and relief | Worn plates print differently |
| Impression setting and press speed | Both change dot gain and transfer |
| Measured L*a*b* per colour-critical patch, with condition and backing | The number the next run is judged against |
| Operator and date, signed | Accountability and traceability |
A supplier that can produce this card for your last approved run can repeat it. A supplier that cannot is going to match by eye against whatever sample is nearest, which is exactly how a warm brown arrives in September. This belongs in your written product specification rather than in an email thread, because email threads do not survive a change of account manager.
Registration is a separate number
Colour tolerance and registration tolerance are different clauses and buyers routinely merge them. Flexo registration on corrugated is coarser than offset; agree a figure in millimetres with the press that will run the job rather than importing a commercial-print number. Then design to survive it: build trap and bleed rather than butting colours; keep fine reverse type off B and C flute, where the flute pitch eats it; and if a design genuinely needs fine detail, that is an argument for a finer flute or a laminated construction, which is a structural decision made before artwork, not after. The flute profile you choose sets a hard ceiling on the print detail the box can hold.
When the box, the hang tag and the sticker must match
This is where colour programmes usually break, and it is a structural problem rather than a quality problem. A carton printed flexo on recycled liner, a hang tag printed offset on coated board, and a label printed digitally on a white face stock are three different processes on three different substrates. They will not produce the same measured L*a*b*, and no amount of pressing the supplier will make them.
What you can do is specify per-substrate targets: one measured reference per component, each derived from the same brand colour, each with its own tolerance, all approved together on one occasion so a human being has confirmed the set looks like a family under a controlled light source. Specify that viewing condition too — a D50 booth, per ISO 3664 — because a set approved under warehouse fluorescents will not hold up under a retailer's LEDs.
There is a real advantage in having those components made on one site: artwork, colour approval and delivery run on one schedule and one master set, instead of three suppliers each matching a different reference at a different time. Jindong Packaging's published CMH listing describes colour-printed retail boxes and gift boxes built to a buyer's artwork with samples in about seven days, alongside colour stickers, hang tags and instruction inserts produced next to the packaging itself — the listing's stated reason being that it keeps artwork, colour and delivery on a single schedule. That is a scheduling and reference-control benefit; it is not a claim that the three substrates will measure alike, and no supplier should offer that.
Common questions
Can a supplier guarantee a Pantone match on brown kraft board?
Not on the brown itself. A PMS reference is defined on white stock; over a brown substrate the printed result shifts and the shift depends on the liner batch. The workable answers are a white-top liner, a white flood underprint, or a measured target recorded on your actual liner and treated as the reference from then on.
My supplier and I measured the same box and got different ΔE values. How?
Almost always a mismatch in measurement condition (M0 versus M1), illuminant, ΔE formula, backing, or the spot measured on a fluted panel. Fix all five in the PO and the disagreement disappears.
The repeat order looks right in the factory photo but wrong on arrival. Why?
Viewing conditions, usually. A phone photo under workshop lighting is not a colour judgement. Specify a D50 viewing booth to ISO 3664 for approval on both sides, and where colour is commercially critical, have it verified during a pre-shipment inspection against the retained master and the recorded values rather than against a photograph.
What to ask the supplier next
Ask these before the first order, not after the second one disappoints. The answers separate a press that controls colour from one that chases it.
- What is your colour-control method? Approved wet proof, drawdown on the actual liner, spectrophotometer with a stated ΔE tolerance, or an eye match to a retained sheet. All four exist in the trade; only the third produces a number.
- Is there a spectrophotometer on the floor, and what measurement condition does it use? M0 and M1 give different answers on OBA-loaded white-top liner.
- Will you issue measured values per run? A short report with L*a*b* per colour-critical patch, condition, backing and number of readings.
- How many retained masters do you keep, how are they stored, and how often are they re-issued? Dark, sealed, three copies, stated interval.
- Which anilox and plate set ran my approved job, and will the repeat use the same ones? Ask for the line screen and BCM in writing.
- How many print units are available, and can you run a white flood underprint? This decides what is achievable on brown liner.
- Are the stickers, hang tags and inserts printed here or bought in, and on what process? It changes what "matching" can mean.
- What is your registration tolerance in millimetres on this flute? Then design trap and minimum type size to it.
- May I see the quality-system certificate and, specifically, the document-control and retained-sample procedures inside it? The certificate on the wall matters far less than whether the procedures behind it actually govern retained masters — read the scope, then ask to see the procedure.
Jindong Packaging is a workable example of how to run that list rather than an example of an answered one. Its published listing covers colour boxes and gift boxes built to artwork from a 12,000 m² plant in Xinqiao, Bao'an, Shenzhen operating since 2007, with two-colour high-speed printer-slotters among the equipment named and board produced in-house, sampling quoted at about seven days and bulk at about fifteen working days from 1,000 pieces. It does not state a colour-control method, a ΔE tolerance or a measurement condition — which is exactly why questions 1 to 3 are the first three, for this supplier and for every other one on the list. Pair the answers with a disciplined sample approval process and the September box stops being a surprise.