Does Virtual Try-On Reduce Returns? What the Data Actually Shows
Does virtual try-on reduce clothing returns? Named sources, Snap's Princess Polly caveat, and how Shopify brands should measure their own reason codes.
Yes, when the refunds are about look and silhouette. Virtual try-on can cut clothing returns by showing the shopper the garment on their own photo before checkout. Antla is Shopify photo try-on for that job. It does not publish a first-party return-rate delta, and it does not pick a size.
The warehouse already processed the dress. The reason code says too small. Finance booked the refund. Nobody asked whether the shopper ever saw the hem on her own legs.
That is the question this page answers: when preview changes the return file, which public numbers you can cite, and which numbers clothing brands should stop repeating.
Returns teams should split look-mismatch from label-mismatch before they credit a try-on app. Editorial image in 1950s film style.
Start with the reason codes, not the app
If size-and-look tickets are a rounding error in your export, skip this page and go fix QC. Preview only pays for itself where the shopper ordered something she could not picture.
Shopify’s 2026 ecommerce returns guide cites the NRF and Happy Returns 2025 landscape: an estimated 19.3% of online sales came back that year, against 15.8% for retail overall. Apparel sits above the online average because fit is hard to judge on a screen. Shopify lists fit or sizing as a primary reason, then mismatch between the item and its photos or copy, then damage, defects, and change of mind. Those are different files. They need different fixes.
Narvar’s 2022 State of Returns still names the fit share: size and fit accounted for 45% of returns that year, up from 42% in 2021 and 38% in 2020. That is 45% of returns, not 45% of shoppers. Treat it as a reason mix, not a conversion slide.
Export 60 to 90 days and sort before you install anything:
| Tagged reason | What it usually is | First fix |
|---|---|---|
| Too small / too big / between sizes | Label guess or a weak chart | Measurements and a size app |
| Didn’t look like the photo / not as expected | Studio shot vs her body | Photography, fit notes, visual try-on |
| Wrong length, too revealing, didn’t suit me | Silhouette she could not simulate | Visual try-on plus model context |
| Ordered two sizes, kept one | Bracketing as a home fitting room | Size tool, then preview on the same PDP |
| Defect, damaged, wrong item shipped | Operations | QC and packing, not a PDP widget |
| Worn once, tags off | Wardrobing | Policy, not try-on |
Wrong-size returns in online fashion is the longer split between a failed label and a failed picture. This page stays on whether preview moves the second pile.
The job try-on can do before a parcel ships
A clothing PDP asks the shopper to mentally paste a sample body onto her own. Size charts answer which label to tap. They do not answer whether the wrap dress will pull at the hip or the blazer will box at the shoulder.
Shopify’s virtual fitting rooms guide puts the commercial job in one line: reduce uncertainty about fit and style before Add to cart. That is more useful than calling the widget engaging. A shopper can spend three minutes on a page and still leave with the wrong mental picture.
Photo try-on is the clothing version of that idea. The shopper uploads one photo in the browser. The app renders your front-facing product image onto that photo. She sees length, coverage, color against her skin, and how the silhouette sits on her frame. She can still pick the wrong size. She is less likely to buy a shape that was never going to work on her.
Narvar’s apparel-returns checklist (updated 8 September 2026) is blunt about the photo gap: 93% of shoppers called product photos important or very important, 72% wanted real-customer images plus more sizing and fit information, and almost half of a 2019 Yotpo fashion survey Narvar cites said the item looked different in person than online. A prettier studio set does not close that file. A self-referenced still might.
Shopify’s returns guide also tells merchants to use 3D and AR where fit drives refunds. Its worked example is Gunner Kennels placing a crate next to a dog, not a dress on a person. Do not import that 5% crate result into a fashion deck. The transferable idea is self-referenced visualization before checkout, without pretending a kennel case study is a clothing result.
Antla virtual try-on does that job from existing product photos. No 3D garment pipeline. Front-facing packshots work; a back-only hero does not.
What it will not do:
- Communicate scratchy wool or clingy jersey in humidity
- Expose cheap lining or loose stitching
- Stop buy-wear-return behavior
- Pick a size, or prove the waist will not pinch
- Rescue a catalog whose top reason code is “arrived damaged”
If those dominate the mix, preview will look like a failed experiment. It wasn’t. You measured the wrong leak.
Named numbers, with the caveats attached
Returns managers get sold round percentages with no denominator. Here is what is actually on the public record.
| Source | Figure | What it measured | Window |
|---|---|---|---|
| Snap AR Enterprise Services | 24% lower return rate | Princess Polly shoppers who used Fit Finder and AR Try-On, versus shoppers who used neither (Snap internal data) | 1 July 2020 to 31 October 2022 |
| Antla first-party returns | None published | There is no confirmed Antla return-rate delta on this site | n/a |
| Antla try-on-user conversion | 3.8% of try-on users purchased | Cohort conversion among people who completed a preview, across 500,000-plus try-ons. Not a return metric. Not store-wide conversion. | Reported on that page |
Read the Snap line the way Snap wrote it. Fit Finder is a size tool. AR Try-On is a preview tool. The 24% compares users of the combined suite with non-users. It is directional for “show the shopper more before she pays.” It is not a clean try-on-only result, and it is not Antla’s number.
People who opt into preview may already be more careful buyers. That selection effect cuts both ways: the cohort can look better than the store average without the widget causing every point of difference. Treat public case studies as a reason to run your own holdout, not as a forecast you paste into a board deck.
The 3.8% figure answers a different question: do try-on users buy? Yes, at a higher rate than a typical Shopify store-wide average, with the caveats on that conversion page. It does not tell you whether those orders come back. Do not multiply 3.8% by an invented return cut. They are not the same math.
This page previously repeated unconfirmed Antla return and conversion claims. Those lines are gone. If a first-party return delta is published later, it belongs here with a date range, a store count, a baseline definition, and the categories in the sample. Until then, your tagged reasons are the evidence.
Try-on is not a size recommendation
Most “wrong size” tickets mix two failures. The shopper tapped the wrong letter, or she tapped the right letter and hated the shape. A size app treats the first. A preview treats the second. Buying one tool for both jobs is how you waste a quarter.
Antla has no size chart or size recommendation of its own. It works alongside any Shopify size app. Keep garment measurements and the size widget next to the selector. Put try-on next to the gallery. “Find my size” and “Try it on” should not share a button.
The longer comparison lives in AI size recommendation vs virtual try-on. I’d start there if your export is still 70% too-small / too-big after the chart is honest. If the file is “looked different” and “not flattering,” start here and leave the chart in place.
A preview can look right in a size that still pinches. That is not a bug in the render. It is a reminder that a still image is not a tape measure.
Categories where a preview is even in the conversation
Impact concentrates where the shopper cannot simulate the garment from a flat photo.
Worth a pilot first: wrap and bias-cut dresses, wide-leg and cropped denim, structured blazers, anything with a high neckline or a short hem, swim and bodysuits where coverage is the whole question.
Usually slower: boxy tees, elastic-waist basics, heavily stretched knits with a forgiving silhouette. Engagement can still rise. Return movement is often small.
Wrong tool: quality defects, color lots that do not match the photo because photography is lying, and wardrobing on occasionwear. Fix the photo or the policy.
Shopify fashion return rate benchmarks is useful context beside your category export. It is not a substitute for it. Two dress brands can share a category and have opposite reason mixes.
Run a 90-day cohort, not a press-release metric
I’d rather see one clean category test than a site-wide toggle you cannot read.
- Baseline four weeks of reason codes on one high-cost family. Split too-small / too-big from not-as-expected, quality, and changed mind. Note bracketing (two adjacent sizes on one order).
- Enable try-on on those SKUs only. Use front-facing product images. Hold button placement constant. A buried control produces a thin cohort and a bored finance meeting.
- Compare try-on users with non-users on the same SKUs for conversion, units per order, bracket rate, and return rate within 30 days of delivery. Review at 60–90 days so late parcels show up.
- Do not mix denominators. A try-on-user return rate is not your store-wide return rate. If 4% of PDP visitors complete a preview, even a strong cohort barely moves the blended number until placement improves.
- Feed the codes back into merchandising weekly. If one dress family keeps coming back as “too short in the torso,” rewrite the length note before you blame the app.
Reduce bracketing orders on Shopify fashion covers the incentive side of two-size baskets. How to build a fashion returns-reduction strategy is the wider loop: PDP, policy, and ops. This page is only the visualization layer.
Plans on antla.io/pricing as of 7 October 2026: Trend $19.99/month (100 try-ons, then $0.16), Runway $49.99/month (500, then $0.12), Unlimited Fashion $199.99/month (2,000, then $0.09). Every plan includes a 7-day free trial. There is no free plan.
Per Antla’s privacy policy: “Images are deleted and not accessible within 72 hours.”
Antla is Built for Shopify. It is not SOC 2 or ISO certified.
Frequently asked questions
Does virtual try-on reduce returns for clothing brands?
It can, when look and silhouette drive the refunds and the shopper sees herself before she pays. Public evidence is directional (Snap’s Princess Polly suite, Shopify’s fit-uncertainty framing). Measure your own try-on users against non-users for 60–90 days. Do not expect a defect-heavy catalog to move.
Has Antla published a return-rate reduction number?
No. Antla has not published a confirmed first-party return-rate delta. The 3.8% figure on the conversion page is try-on-user purchase rate, not returns. Older first-party return-cut language on this URL was unconfirmed and has been removed.
Is Snap’s 24% figure a try-on-only result?
No. Snap reported a 24% lower return rate for Princess Polly shoppers who used Fit Finder and AR Try-On together, versus shoppers who used neither, from 1 July 2020 to 31 October 2022. Fit Finder is a size tool, so the number is not a clean try-on-only result.
Can virtual try-on replace a size chart?
No. Try-on shows how a style looks on the shopper. It does not pick a size. Antla has no size chart or size recommendation. Keep measurements and a size app next to the selector.
How long should a clothing store wait before judging return impact?
Most teams need 60–90 days of cohort data to capture delivery and late returns. A four-week snapshot can show conversion direction. It is too short for returns math.
What happens to the shopper’s photo?
Shoppers upload in the browser. They do not download a shopper app. Antla’s privacy policy states: “Images are deleted and not accessible within 72 hours.”
See the preview on your catalog: Antla.io/demo · Install: apps.shopify.com/antla
About the author: Aaron leads Antla. He would rather a returns manager run a 90-day holdout on one dress family than paste a vendor percentage into a deck.