The short version: Does virtual try-on reduce returns? Probably yes, for the right product categories — but nobody can honestly tell you by how much, and most of the figures you will be shown do not survive being checked. We traced the circulating claims back as far as each one would go. The reductions quoted range from about 20% to 64%, almost all originate from companies selling the technology, and at least one widely-repeated "proof" turns out to be a retailer improvement that the retailer itself attributed to something else entirely. Below: what the claims actually say, where they break, and a method for measuring it on your own catalogue instead of borrowing someone else's number.
Why this guide exists
A retailer asked us a straightforward question: does virtual try-on reduce returns, and by how much? We went looking for a defensible answer and could not find one. What we found instead was a set of confident, precise-sounding percentages that disagree with each other by a factor of three, repeated across dozens of pages, almost none of which say where the number came from.
So this is not another article quoting those figures. It is an attempt to check them.
We should be upfront about our position: we build AR and virtual try-on experiences, so we have an obvious interest in you believing this technology works. That is precisely why we are not going to hand you a number we cannot support. If we quoted you a 40% reduction and you later discovered it traced back to a press release, you would be right to distrust everything else we told you.
The claims, and how far apart they are
Here is the spread of returns-reduction figures currently in circulation, as claims rather than as findings.
Returns reduction attributed to virtual try-on
Each bar is a figure we found asserted in publicly-published marketing or trade content. These are claims we located, not results we verified, and they are shown to illustrate the disagreement — not as evidence of anything.
A three-fold spread is not a measurement. It is a signal that different people are measuring different things — or not measuring at all.
Notice also how the language shifts as the numbers rise. The lower figures tend to be framed as ranges with caveats. The highest tend to arrive as "brands that offer virtual try-on average…", which is a comparison between two groups of retailers rather than a before-and-after within one. Those are completely different claims, and the difference matters enormously, as the next section shows.
The ASOS example, checked properly
ASOS comes up constantly in this discussion, usually as the headline proof. The claim, in various forms, is that ASOS cut its returns rate by roughly 150 to 160 basis points and that virtual try-on was responsible.
The first half is accurate. ASOS did report an underlying returns-rate reduction of roughly 150 basis points year on year in its FY25 results.
So the sequence is: the returns improvement happened, ASOS credited size guides, reviews and a fee policy, and the virtual try-on arrived afterwards. Citing that improvement as evidence for try-on inverts the timeline.
We are not suggesting anyone fabricated this. It is a much more ordinary failure: one article compresses "ASOS reduced returns" and "ASOS is doing virtual try-on" into a single sentence, the next article cites that one, and within a few links the causation has become established fact. It is worth remembering that the fee policy in that list is a genuinely powerful lever on returns behaviour, and one that has nothing to do with any technology at all.
Why this is genuinely hard to measure
In fairness to everyone publishing these figures, isolating the effect is legitimately difficult, and for reasons that are worth understanding before you design your own test.
Self-selection
Shoppers who use try-on are more engaged and more deliberate to begin with. They would likely have returned less anyway. Comparing them to everyone else measures the shopper, not the tool.
Nothing changes alone
Try-on usually ships alongside better imagery, new size guides or a revised returns policy. Any of those can move the number on its own.
Category mixing
Structured tailoring, knitwear and shoes have very different return dynamics. A blended figure across a catalogue hides more than it shows.
The lag
Returns arrive weeks after purchase. Short studies systematically undercount them and flatter the result.
Reporting bias
Deployments that did not work are rarely written up. The published record is the surviving half of the story.
Definition drift
Returns rate by units, by value, or by order? Including exchanges or not? Two honest teams can produce very different numbers from identical data.
What can honestly be said
Strip out everything unsupported and a modest, useful set of statements survives.
How to measure it on your own catalogue
This is the part that actually pays for itself. Rather than adopting somebody else's percentage, generate your own — and design the test so the answer means something.
A rough worth-it calculation
Before running anything, it is worth knowing how big the prize would have to be. Put your own figures in below — nothing is sent anywhere, it calculates in your browser.
What a returns reduction would be worth to you
Rough arithmetic to size the opportunity, not a forecast. It assumes a return costs you the handling and logistics of processing it, and that some returned stock is resold.
The point of this is not the output. It is to establish, before you spend anything, whether a plausible improvement would be material to your business at all. For a lot of retailers the honest answer is that fixing size guides and product copy would move the same number for a fraction of the cost — and that is worth discovering in a spreadsheet rather than after a build.
Where it works and where it does not
| Category | What the shopper is uncertain about | Can try-on answer it? |
|---|---|---|
| Eyewear | How frames suit their face | Yes — this is close to the ideal case |
| Cosmetics | Shade against their skin tone | Largely, though screen colour accuracy is a real limit |
| Jewellery & watches | Scale and proportion on the body | Yes, particularly for size perception |
| Footwear | Appearance yes, fit no | Partly — sizing remains the harder problem |
| Structured clothing | Fit across a specific body | Weakly. This is the hardest case and the one most often oversold |
| Furniture & homeware | Scale in their actual room | Yes, but via AR placement rather than try-on |
The pattern: virtual try-on is strong where the question is how will this look and weak where the question is will this fit. Most of the inflated claims come from applying it to the second question and reporting on the first.
Frequently asked questions
Does virtual try-on reduce returns?
The mechanism is plausible and many large retailers are investing in it, but no publicly available figure is reliable enough to quote as fact. Published claims range from roughly 20% to 64% reduction, almost all originate from companies that sell the technology, and the methodology behind them is rarely disclosed. The effect is also very likely to vary by category — strong for eyewear, cosmetics and jewellery, much weaker for clothing fit. The only number that should influence your budget is one measured on your own catalogue through a controlled test.
Did ASOS reduce returns using virtual try-on?
ASOS did report an underlying returns-rate reduction of roughly 150 basis points year on year in its FY25 results, but it attributed that improvement to upgraded size guides, more customer reviews and changes to its fair-use returns policy including fees for the highest returners, alongside internal AI analysis of return reasons. Its shopper-facing virtual try-on, developed with AIUTA, launched afterwards and rolled out initially to a limited set of products and customers. Citing that returns improvement as evidence for virtual try-on reverses the actual order of events.
Which product categories benefit most from virtual try-on?
Categories where the shopper's uncertainty is about appearance rather than fit. Eyewear is the clearest case, since how frames suit a face is exactly what a visual tool can answer. Cosmetics and jewellery also work well, with the caveat that screen colour accuracy limits shade matching. Clothing is much harder, because fit depends on body measurements a camera cannot reliably infer, and it is the category where returns-reduction claims are most often overstated. For furniture and homeware the useful technology is AR placement at true scale rather than try-on.
How do I test whether virtual try-on works for my shop?
Run a split test at product level rather than shopper level. Enable try-on on one set of SKUs and hold back a matched set with similar category, price and historic return rate. Change nothing else during the test, fix your definitions of returns rate and attribution window in advance, and let the test run for at least one full returns cycle plus a buffer so late returns are captured. Track try-on opens as a separate event so you can tell low usage apart from usage that did not change behaviour, and segment results by category rather than reading a blended average.
Why do published virtual try-on statistics vary so much?
Because they are measuring different things. Some compare a retailer before and after deployment, others compare retailers that offer try-on against those that do not — which mostly measures the difference between the two kinds of retailer rather than the effect of the tool. Definitions also drift: returns by unit or by value, exchanges included or excluded, and different attribution windows all produce different numbers from the same underlying data. Add self-selection by engaged shoppers and the tendency not to publish disappointing results, and a wide spread is exactly what you would expect.
Is virtual try-on worth it if it does not reduce returns?
Possibly, because returns are not the only outcome worth having. Try-on can lift engagement, time on page and confidence at the point of decision, and it can reduce pre-purchase support queries about fit and appearance. It also gives you data about what shoppers were unsure enough to check. Those benefits are easier to measure than returns and arrive sooner, so it is worth deciding in advance which outcome you are actually buying rather than defaulting to the returns argument because it is the one in the sales deck.
We would rather measure it than sell it to you
If you are considering virtual try-on or AR for a UK retail catalogue, we will help you design the test before we quote for the build — including telling you if we think fixing your size guides would move the number more cheaply.
Start with the measurement questionRelated reading: our virtual try-on service, AR product visualisation, e-commerce growth, virtual try-on for UK retailers, AR product viewers versus static photography, and what 3D product models cost.
