Everybody agrees that some product photos promise more than the garment delivers. Almost nobody agrees where the line is, and the usual attempts to place it — how much retouching, how flattering the light, how idealized the model — do not survive contact with real images.
They fail because they measure the wrong thing. A completely unretouched photograph can overpromise badly, and a heavily worked one can promise nothing at all.
Overpromising is not a quantity of retouching
Consider a photograph with no editing whatsoever: one garment, one model, one afternoon. It shows a jacket hanging cleanly on a person whose proportions suit it, under light chosen to flatter the cloth. Nothing was altered. A customer looking at it concludes the jacket hangs that way, which it does — on that person, in that light.
Now consider a heavily composited image where the background was replaced, the crop rebuilt, and the color corrected to match a physical sample. Far more was done to it, and it may state less than the first one did, because none of the work touched anything a buyer was trying to find out.
Whatever separates these two, it is not the amount of intervention. Intervention is the variable everyone reaches for because it is the one you can see in the file.
The test is which question the buyer stopped asking
Here is the definition worth working from. A photo overpromises when it settles a question the buyer had, on evidence it does not contain.
That shifts the test from accuracy to closure. Accuracy requires an external reference — the physical garment, a measurement, a test result — and in most review situations nobody has one. Closure can be assessed from the page itself, by anyone, in a minute: look at an image and ask what a customer would stop wondering about after seeing it, then ask whether the image had grounds to settle that.
Accuracy fails as a working test for a plainer reason than it being hard to define. Whoever reviews the images usually does not have the garment — it is at a supplier, in a warehouse, or not yet made — so the question gets answered by comparing the image to another image, which confirms nothing. Closure needs no sample. It asks what a reader of this page would take away, and the reviewer is already a reader of the page.
Most images settle several things legitimately. A photograph settles color approximately, shape, proportion against a visible body, and what the garment looks like from the angle shown. Those are things light bouncing off an object can carry. The trouble begins with everything else, because an image does not signal which of its apparent answers are supported.
The cost of getting this wrong lands in returns rather than in complaints, and it lands late. It also moves in the opposite direction from the metric most teams optimize, which is its own subject: how product images drive return rates.
Closing a question about something never photographed
The first and largest category is properties that were never in front of the lens.
• Weight. A garment’s mass is invisible, and a well-lit photograph of a heavy coat and a light one can look identical.
• Warmth. Insulation is a function of construction and fiber, none of which reaches a sensor.
• Hand. How cloth feels between the fingers has no optical signature at all.
• Durability. How the garment survives twenty washes is not a property of the afternoon it was photographed.
Each of these is something buyers genuinely want to know, and the image is silent on all of them. The problem is that silence does not read as silence. A confident, well-made photograph reads as a complete account, so a buyer looking at a heavy-looking coat concludes it is heavy, and the conclusion feels like something they observed rather than something they supplied.
The image made no claim. It also did not prevent one from forming, and from the buyer’s side those are indistinguishable.
Answering a general question with one particular case
The second category is subtler and more common in apparel than anywhere else.
Every photograph is of one instance: this garment, from this production batch, on this body, in this light, styled this way, pinned or not pinned. The buyer’s question is almost never about that instance. It is about what will arrive at their house and go on their body.
A photograph cannot mark itself as particular. There is no visual convention for “this is how it fits this person,” so the image reads as how it fits, full stop. When the model’s proportions are near the middle of the range the garment was designed for, the gap is small. When they are not, the image is an accurate record of an outcome very few customers will reproduce, and it will still be read as general.
Using one model across a whole range makes this worse rather than better. Consistency is desirable for other reasons, and it has the side effect of turning a particular body into an implied standard: every garment on the site hangs on the same proportions, so the site as a whole reads as a statement about how the clothes hang rather than about how they hung once. Nothing in any single image is wrong, and the impression is produced by the set.
The same applies to the garment itself. The sample photographed came from somewhere — a first production run, a supplier’s sample, a corrected version — and nothing on the page says which. A photograph of an approved sample and a photograph of an earlier round look the same.
Presenting a proposal as a record
The third category has grown quickly and is the hardest to detect from the page.
Some images document something that existed and was photographed. Others show something plausible that was never in front of a camera — a colorway derived from another colorway, a garment placed on a body it was never on, a detail completed because the original frame did not capture it. Both are image files, both sit in the same grid, and a buyer has no way to tell them apart.
This is not automatically overpromising. A proposal can be accurate, and a record can mislead. It becomes overpromising when the proposal settles a question, because a proposal by definition has no evidence behind the thing it is proposing. The reason it persists is structural: nothing in an ordinary asset library has a field marking which kind an image is, so the distinction is lost within weeks of the shoot — the same gap that makes images outlive the products they describe: what apparel asset management actually maintains.
Where the answers should go instead
Here is the part that keeps this from becoming a counsel of despair. An image that answers nothing is not a safe image; it is an image nobody buys from, and a page nobody converts on settles no questions at all.
The move is not to promise less. It is to answer each question in the field that can carry it and source it:
Weight, fiber, and construction belong in specification fields, taken from the material records rather than written by whoever was drafting copy. Measurements belong in a size chart, with the points of measure stated so the numbers mean something. Fit relative to a body belongs in model information — height, size worn, and any pinning — which converts a particular case into a usable reference instead of a silent generalization. Provenance belongs in a field marking whether an image documents or proposes.
Which to move first is answerable from data you already have. Return reasons are a record of the questions customers got wrong, so the ones appearing most often are the ones your images are closing least reliably. Starting there costs nothing to work out and puts the effort where the evidence already points, instead of on whichever field is easiest to fill.
Each of those takes a question the image was closing without grounds and moves it somewhere with a source behind it. The image is then free to do what images are good at, which is showing what the garment looks like — and the photography step can be briefed against a shorter, more honest list of what it is responsible for: what a product shoot is actually answering.
Questions teams ask about overpromising images
Is a flattering photo automatically an overpromising one?
No, and conflating the two is why this is hard to govern. Flattering describes how appealing the image is; overpromising describes whether it settled something it had no basis to settle. A photograph can be extremely attractive and close no unsupported questions, which is the target rather than a compromise.
How do we audit our own images for this?
Look at each image and write down what a customer would stop wondering about after seeing it. Then check, for each of those, whether the answer came from something in the frame or from something the viewer supplied. The second list is where the exposure sits, and producing it takes minutes per page.
Our model is a standard size. Does that solve the particular-case problem?
It shrinks the gap for customers near that size and leaves it open for everyone else, which in extended ranges is most of them. Stating the model’s height and the size worn converts the image from a silent generalization into a stated reference point. That is a small edit with a disproportionate effect.
Can generated images be used without overpromising?
Yes, when they are not the thing settling a question. A generated image showing styling or context can be entirely appropriate, while a generated image a customer reads as evidence about fit or fabric is not. The distinction is about what the image is being asked to do, not about how it was made.
Who should own this question internally?
Whoever owns returns is the one with the evidence, and whoever owns imagery is the one who can act. Those are usually different people who do not meet, which is why the problem persists in organizations that are perfectly aware of it. Naming a single pair of owners for the review does more than any guideline.
Does adding disclaimers fix it?
Rarely, because a disclaimer sits outside the image and the image is what the buyer processed. What works is putting the real answer in a field the buyer can read as an answer — a measurement, a weight, a stated size worn — rather than negating the impression the photograph already created.
Where this leaves you
An overpromising photo is not a dishonest one, and usually not a heavily edited one. It is an ordinary, well-made image that let a buyer stop asking something it had no grounds to settle: how heavy the cloth is, how it sits on a different body, whether the thing shown was ever photographed at all. The useful review question is not whether an image is accurate but which question it closed. Then move the closed questions that lack evidence into fields that have some, and let the photograph carry what light can actually carry.
List what each image closes
Take one product page and write down, for every image on it, what a customer would stop wondering about after looking at it. Mark each entry according to whether the answer came from something visible in the frame or from something the viewer filled in. The second group is your exposure. For each of those, decide which field should carry the answer instead — a measurement, a stated weight, the model’s size worn — and check whether that field currently has a source behind it.
Written by
