How Color Matching Works in Clothing Manufacturing

Three deliveries of the same dress, in the same colourway, arrive eight weeks apart. Piece one is the colour the buyer signed off. Piece two is close enough that nobody says anything. Piece three reads a shade deeper in the studio, and when all three are laid side by side on a table, a customer can see it in a phone photo.

Nothing was substituted. The dye recipe was not changed. Every garment passed inspection against the approved standard, because the approved standard did not describe the colour precisely enough to fail anything.

That is the pattern behind almost every "the colour is wrong" conversation in garment production: not a mistake, but an approval that was narrower than the thing it was approving.

This article works through what actually moves a shade, how a colour moves from a reference to a bulk approval, what the numbers and the light actually tell you, and how to close the gap before bulk is booked.

What actually decides a bulk shade

A single swatch approval settles one physical sample, viewed once, under one light. It does not fix the production colour, because four things move it afterwards: the dye lot the fabric was dyed in, the fibre and construction being dyed, the finishing the fabric goes through, and the light under which someone looks at it.

There is no fixed ranking among those four. Which one dominates depends on the fabric, the colour, the mill and the customer — a deep saturated shade on a filament fabric behaves differently from a pale shade on a spun one, and a finish that moves a black will barely register on a beige. Treating them as a shortlist of suspects rather than a hierarchy is what makes a diagnosis possible. It is also why checklists that assert "dye lot matters most, lighting matters least" are wrong as often as they are right.

The work of colour control is therefore not "choose the colour". It is agreeing a reference, a tolerance and a method of comparison — and then checking bulk against all three, in the finished state.

The four variables that move a shade

1. Dye lot

Fabric is dyed in batches. Each batch — a lot — is a separate chemical event on separate fibre, and two lots dyed to the same recipe will not be identical. They can be very close, and "very close" is a real, measurable standard. It is simply not identical.

This is why a supplier asks how many rolls or kilos you need before quoting a colour: the answer decides whether the order sits in one lot or spans two. A single-lot order is a genuinely different colour risk from a multi-lot order, and it is worth knowing which one you are buying before you buy it.

2. Fibre and construction

The same recipe on two different substrates does not produce the same visual colour. Fibre type changes how dye is absorbed and how much light the surface throws back. Construction changes the surface: a smooth satin and a textured crepe can sit at the same measured colour value and still read differently to the eye, because the eye is reading reflected light, not a number.

This is the technical reason a colour approved on one fabric cannot simply be "transferred" to another. Two fabrics that look like variations of each other on a card can be far apart once they are dyed. If the substrate is changing, that is not a colour question — it is a fabric question, and fabric weight and behaviour explains why a change in the cloth is a change to the garment regardless of whether the shade follows.

3. Finishing

Finishing happens after dyeing and it can move a shade. Washing, softening, calendering, heat setting and any surface treatment change the way the cloth reflects light. A lot can measure correctly at the dyehouse and read differently after finishing — which is why colour control has to include the finished state, not the dyed state alone.

This is one of the most common sources of a genuine, unintended mismatch: the approval was taken on greige or dyed fabric, and the garment the customer receives has been through a process nobody sampled.

4. The light you are looking under

Two pieces of fabric can match under one light source and differ under another. This is metamerism, and it is why "it matched in the office" and "it doesn't match in the shop" can both be true. Daylight, warm indoor light, cool retail lighting and a phone flash all render the same cloth differently.

The practical consequence is simple and frequently ignored: colour must be judged under an agreed light, and the reference sample has to be stored in conditions that keep it valid. A reference swatch left on a windowsill for six weeks is no longer a reference.

How a colour moves from reference to approval

The reason "we approved the colour" is such a weak statement is that approval is a sequence, not an event. Each stage has an output, and a stage that is skipped is a question that stays open.

Two stages in that list are routinely compressed or skipped, and both cause problems later.

Lab dips on paper are not lab dips. A dye recipe sheet, a printed colour card or a digital rendering is not a dyed sample. A lab dip is a small dyeing carried out on the substrate — or a representative substrate of the same fibre and construction — because the whole point is to see how the recipe behaves on that material. Approving colour from a printout is approving a document about a colour rather than a colour.

The finished-fabric check is the one that gets dropped. Steps 1 to 7 are about getting the dye right. Step 8 is about what happens to the dye afterwards, and it is the step that catches the case nobody expected.

CIELAB and change in E — what a buyer actually needs

Instrumental colour measurement is worth understanding at the level that changes decisions, and no further. Three coordinates and one difference figure are enough.

Four practical points, and they are the ones that matter commercially.

dE alone is not a description. A single number tells you a difference exists and how large it is. It does not tell you whether the bulk went darker, or shifted greener, or both — and those have different remedies. Ask for the components (dL*, da*, db*) alongside the summary figure, because the direction is what tells you whether to adjust the recipe or the process.

There is no universal dE tolerance. There is no figure that is correct for every fabric, every colour, every market and every brand — the same way there is no universal weight tolerance in a fabric specification. A deep saturated shade can tolerate a larger numeric difference before the eye registers it; a pale neutral can look wrong at a much smaller one. The number has to be agreed between buyer and mill, for that colour on that substrate, and written down.

The formula matters as much as the number. Different colour-difference formulae applied to the same pair of samples return different figures, and the difference between them is not small enough to ignore when a tolerance is tight. So a specification that says "dE under 1.0" without naming the formula is not a specification yet. Expect to name the formula, the illuminant and the observer — for example, a specified illuminant at a stated observer angle — alongside the tolerance.

Instrument and eye each do part of the job. The instrument gives you an objective, repeatable number and a direction; the eye, under the agreed light, tells you what a customer will see, including the metamerism the instrument cannot see. Most production standards use the instrument as a first filter and the visual check as the confirmation. The measurement and difference-calculation methods themselves are formalised in standards such as ISO 105-J03, AATCC EP7, ASTM E308 and ASTM D2244 — they define how to measure and how to express a difference, not what your tolerance should be.

Shade band and the continuity card

Here is the expectation that has to be corrected before any of the rest works: bulk fabric is not trying to be laboratory-identical to the approved sample. It is trying to sit inside an approved range.

That range is the shade band. A shade band is a physical card — swatches, not a numeric range on its own — showing the approved centre on the approved substrate, plus the light and dark limits that have been agreed as acceptable around it. A production lot whose swatch sits inside the band is acceptable. A lot that sits outside it is not, whatever the number says.

A continuity card does the adjacent job: it records where each lot sits relative to the band, so the next production run — new lot, new dyeing, possibly a new season — can be matched to the same position rather than renegotiated from scratch. That single record is what makes a colourway stable across seasons rather than merely approved once.

Three practical notes:

• Hold the band physically, not numerically. A written range cannot be compared against a roll at the delivery door, and the band's stored condition is part of its validity.
• Store the band and the reference away from light and heat, labelled with the recipe and the date. An unlabelled, sun-faded swatch is worse than no swatch, because it produces confident wrong answers.
• Aim at the centre, not at the edge. A lot deliberately produced at the limit of the band has left no room for the next lot, and repeat orders will drift out of the range one step at a time.

Solid dye and print are not the same approval

Everything above describes a solid-dyed fabric: the colour is in the fibre, and the approval process is about recipe, lot and finish.

Prints work differently, and the vocabulary differs with them.

• You do not lab-dip a print. A lab dip tests how a dye recipe behaves on a substrate. In print, the colour is carried by the paste and the screen, so the equivalent question is whether the artwork reproduces correctly on the production cloth.
• The trial is a strike-off. A strike-off is a trial print of the artwork on the target base cloth. It is where registration, colour, coverage, the hand of the printed area and the repeat all get checked — because a print can be the right colour and still be wrong: the repeats can misalign, the paste can sit differently on a textured base, the coverage can be patchy where the cloth is uneven, and the printed area can stiffen the cloth in a way the design did not anticipate.
• Approval runs artwork, screen and print as one chain. The artwork is prepared, screens or engravings are made, a strike-off is produced, corrections are made, and the corrected version is approved. A change at any point in that chain — a colourway, a repeat size, a base cloth — invalidates the strike-off and requires a new one.
• The shades still behave afterwards. A print can shift under finishing and can be more sensitive to light than a solid dye, particularly in pale or highly saturated colours. The finished-state check applies to prints too.

So the first question about a colour problem is not "why is the shade off". It is "is this a dyed colour or a printed one" — because the two have different approval documents, different failure modes and different fixes. The two sit at the same stage of the programme, which is why colour and print development both belong to fabric sourcing rather than to the sewing floor.

Diagnostic tree: reading the symptom

When a colour problem appears, most disputes are settled by identifying which row you are actually in — and several rows are nobody's fault, which is itself worth establishing early.

Two rows are worth pulling out.

Shading is caught along the roll, not at the corner. A slow variation in shade down the length of a roll is invisible if you look at a folded end. This is one reason fabric inspection happens on receipt, before cutting: a shading issue identified on the roll costs a conversation, while the same issue identified in finished garments costs the garments.

The component row surprises people most. Colour approval is usually conducted on the main fabric, while linings, trims, thread, zips and bindings are approved quietly or not at all — and then photographed together. Approving components as a set, against the same reference and under the same light, costs one meeting.

How colour is actually held in production

Colour control on a production floor is not one decision. It is a chain of them, and every link has to hold.

Lab dips before the order. Before bulk dyeing, the dyer produces trial dyeings to proposed recipes and they are submitted for approval. This is where the recipe is settled, and where a buyer gets the first realistic view of the shade on the actual substrate rather than on a screen.

Lot planning. Rolls are planned so that a single order consumes as few lots as possible. Where an order must span lots, the splits are recorded and pieces ship knowing which lot each one came from. That record is what turns "the colour varies" from an argument into a fact with a cause.

Shading and selvedge checks. Shading — a slow variation down a roll — is caught by inspecting along the length, not by looking at a folded corner. It is part of why fabric inspection happens on receipt, before cutting. The wider programme that check belongs to is set out in what a brand should check in garment quality control.

Cutting discipline. Where an order spans lots, a marker can be laid so that all pieces of a single garment come from one lot. It costs a little fabric and removes the worst version of the problem: one sleeve from lot A and one from lot B.

Finishing trials. Where a finish is known to move shade, the shift is measured on a trial piece before bulk goes through, so the effect is known rather than discovered.

Reference storage. Approved swatches, bands and continuity cards are stored away from light and heat, labelled with the recipe and the date, and re-baselined before each production run.

Why "we approved the swatch" is not "the colour is fixed"

An approved swatch is a physical object. It cannot describe the population it is meant to represent unless three things are attached to it in writing.

1. A reference and a tolerance. Not just "this is the colour", but "this is the reference, and the bulk may sit within this agreed range around it" — expressed as a shade band, a difference figure, or both, with the formula, illuminant and observer named.
2. A comparison method. Measured instrumentally, judged visually, or both. Under which light. Against which standard. By whom.
3. A scope and a condition. Which components the approval covers — main fabric only, or fabric plus lining, trims and thread — and whether the reference is dyed-state or finished-state.

Without those three, the swatch is a sample rather than a specification, and every later conversation becomes a matter of opinion.

Where colour sits in the overall specification is covered alongside everything else that belongs in the tech pack, and it is worth reading what a fashion tech pack should include before writing the colour section of your own. The upstream question — whether the substrate itself is right — sits in how to choose fabric for a dress design, and satin, crepe and chiffon is a useful reminder that substrate is part of the colour specification rather than a separate decision.

How to write a colour specification

A colour specification is short — five lines — and it is the difference between approving a colour and approving a piece of fabric.

None of the five is expensive. Four of them are sentences. They go missing because they have to be written before anyone is worried about colour, at the point where everyone is still excited about the design.

What the mill needs from you

The dyehouse cannot hold a colour it was not given. In practice it needs the substrate it is dyeing, the reference and the tolerance, whether the target is the dyed or the finished state, and how many metres are being booked — because that answer determines whether the order spans one lot or several.

That last item is where buyers are most often surprised. The lot plan is a direct consequence of order size and it is available before production starts. Asking for it at quotation is normal and reasonable; discovering it after delivery is what turns a mild variation into a claim.

What you need from the mill

Symmetrically, ask for four things in return:

• The lot plan for your order, and where the splits fall if there is more than one lot.
• Lab dips, or a strike-off for prints, on the actual or representative substrate — not a digital rendering of a colour.
• The finishing route, and whether the finish is known to move the shade.
• The reference condition the bulk will be measured against, so your comparison and theirs use the same baseline.

If a mill cannot answer the first and the last, the colour is not yet controlled, whatever the swatch looks like.

The pre-bulk colour approval card

Everything in this article reduces to one page to be completed before bulk is booked — the point at which colour is still free to fix. After bulk, the only available moves are expensive ones: re-dye, sort, or accept.

Eight lines, one meeting, and colour stops being a recurring argument. Skip it and the same complaint returns every season, from a different angle, on a different garment — with nobody able to say which of the four variables moved.

If you want colour settled against a written reference and a stored band rather than a photograph, send us your colour reference and target components and we will confirm the reference, the tolerance, the condition and the lot plan before anything is dyed. Fabric, linings and trims are developed and approved as a set under fabric sourcing; the inspection side, including where a colour check sits in a sampled inspection, is described under quality control and in what AQL inspection governs.