Do virtual try-ons reduce returns?
Every vendor selling a try-on tool, us included, has an obvious incentive to tell you it cuts returns. Most of them quote a percentage. Follow those percentages back and they usually lead to a blog post citing another blog post citing a vendor’s own case study with no methodology attached.
- Published
- Reading time
- 5 minutes
US return rate, share of sales, 2023
Bought online
17.6%
All retail
14.5%
Bought in store
10.02%
Source: National Retail Federation, 2023 Consumer Returns in the Retail Industry, with Appriss Retail. “Bought in store” is the report’s pure brick-and-mortar rate.
So this post does something less satisfying and more useful: it sets out the return-rate data that comes from a named, checkable source, says plainly what that data does and does not establish, and explains why we are not publishing a figure of our own.
What the sourced data actually says
The most recent returns study from the National Retail Federation that we could locate on their site is [2023 Consumer Returns in the Retail Industry](https://nrf.com/research/2023-consumer-returns-retail-industry), published 22 December 2023 in partnership with Appriss Retail.
| Measure | 2023 figure |
|---|---|
| Total US merchandise returns | $743 billion |
| Overall return rate, all retail | 14.5% of sales |
| Online return rate | 17.6% ($247 billion) |
| Pure brick-and-mortar return rate | 10.02% ($371 billion) |
| Returns per $1bn of sales, average retailer | $145 million |
Two things in that table are worth sitting with.
Online returns run well above in-store returns — 17.6% against 10.02%. The gap is the price of buying something you have not seen in person.
The absolute numbers are large enough that small movements matter. At $145 million of returns per billion in sales, a change of even a point or two in the rate is a serious amount of money for a business of any size.
A note on currency: these are 2023 figures, and we are citing them as 2023 figures. If you have seen a more recent NRF online return rate quoted elsewhere, check whether the source links to an NRF page that exists. Several widely-repeated numbers in this category do not.
What that data does not establish
The NRF figures describe the size of the returns problem. They say nothing whatsoever about whether virtual try-on fixes it. No part of that study tested a try-on tool.
That distinction gets blurred constantly in marketing material: a real statistic about returns is placed next to a claim about try-on, and the reader is left to join them. We would rather not do that.
What can be said reasonably, and no more:
- Online return rates are substantially higher than in-store rates.
- Buyers cannot assess fit, colour or finish online the way they can in a shop.
- A preview on the customer’s own photo addresses that specific gap.
That is a plausible mechanism, not a measured outcome.
Why we publish no figure of our own
We have not run a controlled study of return rates in live stores using Softbox. We would need meaningful volume across several merchants, a proper comparison group, and enough time for returns to actually come back. We have not done that work.
Publishing a number anyway would be easy and would undermine everything else we have written. We have been equally explicit elsewhere: we did not benchmark generation speed on phones, and we do not claim a conversion lift. What we did test is an engineering comparison on identical inputs, which is published in full in what we learned building a product preview tool.
If we run a returns study later, we will publish the methodology with it, including the stores that did not improve.
What to measure in your own store
You do not need an industry average. You have the only data that matters about your own customers, and a try-on feature is straightforward to evaluate honestly.
Start with your baseline, by reason. Most platforms capture a return reason. Separate “wrong size or fit”, “colour not as expected” and “changed my mind”. Only the first two are things a preview could plausibly affect; the third is not try-on’s job.
Roll out partially, not everywhere. Enable previews on one category and leave a comparable category alone. Comparing two periods is weak — seasonality and product mix move return rates on their own. Comparing two categories over the same period is better.
Wait for the returns window. If you accept returns for 30 days, data from the first fortnight tells you nothing.
Check the preview is honest first. A preview that flatters the product sets an expectation the parcel cannot meet, which moves returns the wrong way. Whether a model changed anything it should not have was the main test in our side-by-side model comparison.
Watch return rate among customers who used it. Shoppers who use a preview are self-selecting — more engaged, more considered — so they may return less for reasons unrelated to the tool. That caveat is exactly the one vendor case studies tend to omit.
Count the previews, not just the returns. Usage tells you whether the feature is being found at all. A feature nobody uses cannot change anything, and that is a placement problem rather than a model problem.
The honest summary
Online returns are high and expensive; that part is documented. Whether a virtual try-on moves your rate depends on what you sell, why your customers return things, and how visible you make the feature — and the only reliable evidence will come from your own store.
The case for trying it is not a borrowed statistic. It is that the customer is currently guessing, and a preview on their own photo replaces the guess. Whether that shows up in your returns is a question your data can answer and ours cannot.
Questions people ask
Is there a reliable figure for how much virtual try-on cuts returns?
Not one we can point to from a named primary source with a published methodology. Figures circulate widely; most trace back to vendor material.
What return rate should I expect online?
The NRF put the 2023 US online rate at 17.6% across all retail. Your category matters enormously — apparel typically runs higher than average, homeware lower.
Could a preview increase returns?
In principle, if it set an inaccurate expectation. That is why “did it change anything it should not have” was our main test when comparing models — a preview that restyles the customer’s outfit is making a promise the parcel will not keep.
How long before I can tell?
At least one full returns window after a meaningful number of previews, and longer if your volume is modest.
Does this apply to wall film and car wraps?
Returns behave differently for made-to-order and cut-to-length goods, where the relevant outcome is often a cancelled or disputed order rather than a return. Measure the thing that costs you.
Where do I start?
Baseline your return reasons first, then enable previews on one category. The practical setup is in adding previews to BigCommerce.
