Returns are designed, not marketed away

A high return rate can expose a product or information problem that marketing and customer service cannot resolve alone. Those teams can improve the page, explain fit and report what customers say; product, production and fulfilment teams must also act where the garment or delivery has failed. Start with the reason for the return rather than assigning the whole rate to one function.

The causes sit upstream and they are ordinary. Fit that is not what the size chart implied. Colour that does not match the photograph. Fabric weight or handle the product page never conveyed. A size chart written in a way the customer could not act on. A delivery slow enough that the reason for the order had passed by the time the parcel arrived. Each of those is a decision taken months before the parcel shipped, by somebody who could have decided otherwise.

01Fit that is not what the chart implied

Fit approval is useful returns-prevention work, and it needs to cover the offered range. A fit session judges a sample on a body and signs off the measurements bulk will be made to. Two things go wrong there. The garment is approved on a fit model whose proportions are not the customer’s, so the entire range grades off a shape nobody buys. Or approval happens on the base size alone, and the sizes furthest from it are never seen on a body before they are made. Fitting the top and bottom of the range costs one more sample round and answers a question the returns data will otherwise ask at full price.

Grading is the mechanism underneath the second of those. A graded size range is one approved garment with a set of rules applied to it, stepping each measurement up and down by an agreed amount. Bodies do not change uniformly across a size range: the distance between one size and the next is not the same at the chest as it is at the neck, the armhole or the sleeve, and it changes shape as well as scale at the ends of the range. A grade that steps everything by the same logic produces a garment that is correct in the middle and progressively wrong at both edges, which is why the returns concentrate at the extremes while the base size behaves.

There is a second, quieter mechanism that has nothing to do with the pattern. Fit descriptions are relative, and the reference point belongs to the customer rather than the brand. Regular, relaxed and slim describe different garments in different wardrobes, so a customer who orders the size that worked in a shirt they already own is measuring against that shirt, not against the block this one was cut from. A fit that is internally consistent can still be a size out for most of the people arriving from a category where the word means something else.

Then measure what actually shipped. A specification states what a garment should measure. A production sample and a check across the delivered cartons state what it does measure. Tolerance is real, and a style that lands consistently at one edge of its tolerance will read as a size out to a share of the people who buy it. That check is also the only way to tell a pattern fault from a production fault, and the two have completely different fixes: one is a new block, the other is a conversation about the goods that were delivered.

02A size chart the customer could not use

The size chart is the other half of the problem, and it is usually the weaker half. Two different charts exist. A body chart says what body a size is meant to fit. A garment chart says what the garment measures flat. Publishing one without saying which is why a customer measures a favourite shirt, compares it against a set of body measurements, and orders the wrong size. Publish the garment measurements, name the points measured, explain how to measure them, and keep the body chart alongside where it helps.

Some customers compare a garment they already own; others measure their body or use a familiar size as a starting point. A body chart and a flat garment chart answer different questions. Label each clearly, show where the tape goes and explain the ease built into the garment so customers can use the method that suits them.

Naming the points is not pedantry either. Chest measured flat across the underarm seams and chest measured at the fullest point are different numbers on the same garment. Length taken from the highest point of the shoulder and length taken from the collar seam differ by the depth of the neck. Two brands can publish the same figure for the same word and mean two different distances, so a chart without a diagram or a sentence saying where the tape goes transfers the ambiguity straight to the customer.

One chart across a whole range is the other common failure. A boxy overshirt and a fitted shirt cut from different blocks do not share a garment chart, and a single brand-wide table is wrong for at least one of them. The chart belongs to the style. Where a style is offered in more than one length or leg, each variant needs its own row rather than a footnote.

03What a photograph does not carry

Colour is the failure that arrives with the parcel opened, and it has two separate halves. The first is capture. An image is a record of the light that fell on the garment, so the source decides the record: daylight, tungsten and various LED fixtures each contain a different spread of the spectrum, and the same garment photographed under two setups is two colours in the file. White balance taken off a neutral reference makes the record honest; taken off the garment itself it makes the garment neutral by definition, which is exactly the wrong outcome. Light bounced off a coloured wall or floor tints the cloth. Retouching towards a livelier image moves saturation away from the goods. And because the model shot and the flat shot are usually taken in different sessions, one colourway frequently appears twice on the same page in two versions, leaving the customer to decide which one is true.

The second half is display, and it is not controllable. The screen the customer is looking at is uncalibrated, set to a colour temperature somebody chose once, running an adaptive brightness that responds to the room, and possibly in a warm night mode. No amount of care at the shoot reaches that far. What care does reach is that the file is right, that every image of one garment agrees with every other image of it, and that the colour is described in words as well as shown, because a written description survives a screen and an image does not.

There is a third path to the same complaint that is not photographic at all. Cloth is dyed in batches, and batches move, so a garment photographed from one lot and shipped from a later one can leave an image that was accurate at the time and a parcel that does not match it. Checking the live images against a garment from the lot that is actually shipping is the only thing that catches it, and it catches a colour drift in the goods at the same time.

Weight, handle and opacity fail differently, because a photograph does not carry them at all. An image records colour, shape and, weakly, surface. It does not record mass, drape, thickness, stretch, recovery, warmth, or whether light passes through the cloth. Those are precisely the properties a customer evaluates in a shop by taking the garment in their hands, and online they are replaced by inference from the picture and from the price. Where that inference is wrong the parcel is a surprise, and surprise is the feeling behind most of what gets recorded as poor quality even when the construction is sound.

Opacity is the quietest of them, because it fails after the sale rather than at the moment of opening. A white shirt, a pale trouser or a fine knit can look correct on a hanger and be transparent in daylight, and the customer discovers this on the first wear rather than in the parcel. Weight decides whether a shirt hangs or clings. Handle decides whether the cloth feels like what it cost. Recovery decides whether a knit still has its shape by the afternoon, and wash behaviour decides whether the second wear is the same garment as the first.

The product page is a specification the customer reads. Give it the things returns are usually about: the fabric and its weight, whether it is opaque, whether it holds shape or relaxes through the day, how it behaves after washing, and what the fit is meant to be described in plain terms instead of adjectives. Show the garment on more than one body where that is possible, and state what each person is wearing and what they measure. Photograph colour under consistent, calibrated conditions and check the final images against the garment that will ship. A close image of the surface answers handle better than a paragraph, and a short piece of movement answers drape better than any still.

None of that is persuasion. It is documentation, and it operates at the point where a customer either buys one size or buys two with the intention of sending one back. A page that says nothing specific does not avoid the risk; it moves the guess onto the customer, who is guessing with less information than anyone else in the chain.

04Delivery slow enough that the intent decayed

A purchase is made in a state of intent, and intent has a half life. The garment was bought for something: an occasion, a change in the weather, a trip, a replacement for something worn out. Delivery time runs against that reason. When the parcel lands after the event, after the season turned, or after the customer found something else in the meantime, the garment is no longer being judged against the need that produced the order. It is being judged against no need at all, and it goes back.

What makes this cause hard to see is that it disguises itself in the data. The reason recorded is almost never slow delivery. It is changed my mind, or no longer needed, or ordered by mistake, all of which read as customer behaviour and are in fact a logistics outcome. A cluster of those reasons concentrated on the orders with the longest transit time is the signal, and it only appears if the transit time is being kept alongside the return reason.

Waiting also raises the bar the garment has to clear. Anticipation accumulates across a long wait and an ordinary garment arriving at the end of one underperforms against the version of it the customer has been imagining. Split deliveries do something similar for a different reason: two items bought as an outfit and arriving apart get judged separately, and the one that arrives alone has lost the thing it was bought to go with.

The lever here is the promise rather than the speed. A stated date that the operation can hold behaves better than a faster date that is missed, because a missed date turns the delivery into a complaint before the garment has been seen, and a customer who is already annoyed opens the parcel looking for confirmation. Where the wait is structural, as with a pre order or a made to order run, the antidote is contact during it: the intent has to be maintained rather than left alone to decay.

05A return is evidence only if the reason survives the dropdown

Which makes returns the clearest feedback available about the specification. They arrive with a reason attached, from people who wanted the product enough to pay for it, and they are patterned.

The patterns are legible once the reasons are kept. Too small concentrated in one or two sizes points at grading, the rules that step measurements up and down from the size the garment was approved on. Too small spread evenly across the whole size range points at the base size itself, or at a block that does not suit the body the brand is actually selling to. Not as pictured points at colour management or at photography. Poor quality is frequently not a construction fault at all: it is a customer whose expectation of weight and handle was set by an image and contradicted by the garment.

That only works if the reason survives in a usable form. A dropdown on its own flattens the signal, because someone returning for fit will pick whichever of four options is closest. The free text box is where the specific sentence appears: the sleeve was long, it was fine until it was washed, the colour is browner than the picture. In a first season those can all be read individually, and they should be.

Keep useful context alongside each return: whether several sizes were ordered to compare, whether an exchange followed, whether the garment was worn or washed, its colourway and its production lot where known. Bracketing and exchanges may point to unclear guidance, personal preference or fit problems; neither proves a grading fault. Compare those reports with the specification and returned garment before changing the pattern.

The reason also has to attach to the goods and not only to the order. A file of returns by customer answers questions about customers. A file of returns by style, size and colourway answers questions about the range, which is where the decisions live. And the returned garment is itself an artefact worth keeping: it can be measured, compared against the specification it was made to and against the sample that was approved, and carried physically into the next fit session, where it does more work than any summary of it. Reviews and customer service transcripts carry the same evidence in more detail, and usually nobody owns them.

06The freight is not the loss, the repeat purchase is

The cost structure argues the same way. A return is paid for in the outbound shipping, the inbound shipping, the handling and inspection, the repack, the working capital tied up while the garment travels, and the markdown when it cannot go back to full price. It also holds stock out of the selling window for weeks. All of that is spent after the decision that caused it, on a garment already made.

Stock distortion is the effect underneath that. A garment that comes back re enters the stock file weeks later, in a size that has often already been reordered or already gone to markdown, so the system reports availability in a size whose selling window has closed. Running the other way, the original order counted as demand at the moment it was placed, so the buy for the next season is built on a number that includes garments which came home. The sizes that return most are therefore the sizes most likely to be over bought next time, which is the same fault compounding rather than correcting.

A further cost can be the purchase that never happens afterwards. A customer who returns has spent their own time on a parcel that failed a promise, and even where the process was faultless the thing they learned is that this brand’s sizing cannot be relied on. That knowledge is a reason not to order again, or to order again only with a second size in the basket, which makes the next transaction worse than the first. The first sale to a customer is the one that costs the most to win; the second is the one that pays for winning them. A return does not just cost the freight on one parcel. It can put the repeat order at risk; good resolution can also preserve the relationship.

07A high return rate can be structural, which is not the same as unavoidable

In some categories the rate is high because of what the category is. Where the garment is closely fitted and the body is the variable, where the customer genuinely cannot know their size without putting it on, where the purchase is an occasion piece chosen against a deadline, buying more than one and sending one back is how the category is shopped. That behaviour is not a fault to be engineered away; it is a cost of trading in that product and belongs in the operating plan from the start. The rate on its own does not tell those two situations apart, which is why the rate on its own is close to useless as a diagnostic.

Three comparisons can guide the investigation. Check whether returns concentrate in a particular style, size, colour or delivery; whether an exchange or retained item suggests the customer was choosing between options; and whether the pattern changed after a new lot, maker, photograph, chart or carrier. None is conclusive on its own. Use the number sold as the denominator, allow for the return window and read the actual comments before attributing a cause.

The strongest objection is that returns are part of selling clothing online. A customer cannot fully assess every garment before buying it, so some returns belong in the operating plan. Return charges, windows and service choices can affect both conversion and cost, but the effect needs to be measured for the brand. The policy must also preserve the customer rights that apply in each selling market.

All of which is an argument for offering returns, and none of it is an argument for not knowing why they happen. A lawful, workable policy is the starting point. The rate can still change: it moves in response to the fit approval, the grading, the chart, the images, the words on the page and the delivery promise, all of which are decisions somebody makes. Treating the whole of it as a fixed cost of trading means paying the structural part, which is unavoidable, and the designed part, which is not, and never being able to tell which is which. Accepting the cost and reading the causes were never alternatives.

Absorbed as a cost, returns keep arriving for the same reasons. Read as evidence, they change the next range: the fit approval, the size chart and the product pages get written against what customers reported instead of what the brand assumed. The work that removes a return is done long before the garment is packed, on a fit model, in a specification, and in what the page is willing to publish.