Does Virtual Try-On Actually Reduce Returns or Just Increase Add-to-Cart?

Every deck in this category leads with the same slide. Conversion up 27 percent. Add-to-cart up 35 percent. It is real, and it is the number every merchant wants to see. But if you stop reading the deck after that slide, you will make a decision your P&L does not agree with.

Here is the part nobody puts on the slide. Shopify’s 2024 merchant survey found fashion brands using AR try-on saw an average 27 percent lift in conversion versus static catalogs. Early adopter data from Genlook shows shoppers who actually generate a try-on image convert to add-to-cart 35 percent more often than those who open the widget and do not use it. Those numbers are good. Then you look at what happens on the return side. Forrester’s 2025 analysis of retail technology ROI found return rate reductions averaging 23 percent across apparel categories, with some products hitting 40 percent and others near zero.

Twenty seven to thirty five percent on conversion. Twenty three percent on returns, and that is the average, not the floor. That gap is not noise. It is telling you something specific about what virtual try-on actually does.

Confidence to Order Is Not the Same as Confidence in Fit

Add-to-cart is a decision metric. A shopper saw the garment on a body, felt good enough about it, and moved forward. That is a real signal and it is worth money. But it does not tell you whether that shopper is confident the garment will fit, or just confident enough to try.

Return rates are where the truth shows up. The ICSC’s 2024 survey put online apparel returns at 22 percent, against 6.2 percent in-store. That gap is not about whether the customer wanted to buy. It is about what happens after the box arrives. In a store, you try it on before you pay. Online, you predict, and then you find out.

Virtual try-on is prediction technology. It shows what a garment might look like on a customer’s body. Showing is not guaranteeing. Fabric behaves differently worn than visualized. Seam placement, stretch, drape, none of that fully survives a two-dimensional rendering. A shopper can see a great fit on screen and still be surprised when the actual item shows up.

Bracketing Does Not Care How Good Your Visualization Is

The bigger obstacle is not the technology. It is behavior. Gen Z shoppers lead bracketing behavior at 51 percent participation, averaging 7.3 returns per shopper against a $174 average order value. Bracketing means ordering multiple sizes or colors of the same item with the plan to keep one and send the rest back.

Virtual try-on changes the moment before add-to-cart. It does not change the math behind bracketing. If a shopper knows there is a 10 percent size variance between two brands, they will still order two sizes to be safe, try-on or not. The widget might make them slightly more confident going in, but it does not remove the variance that made them bracket in the first place. That shopper still generates two orders and one return. The add-to-cart number looks great. The return rate barely moves.

Where the Return Reduction Actually Comes From

There is a real pattern underneath the averages. Return reduction happens when the technology moves from showing what a garment might look like to predicting what it will actually measure against a specific body. Visualization and fit prediction are not the same product, even when they look identical on a product page.

The 23 percent average Forrester tracked is most likely being pulled up by implementations that went past pure visualization into actual dimensional mapping. Chest width, sleeve length, waist gap; these are measurable variables, not impressions. A system that maps a customer’s real dimensions against the garment’s actual construction produces a different outcome than one that just renders the garment on a realistic-looking body.

That distinction shows up hardest in specific categories:

• Swimwear and lingerie sit at 30 to 35 percent return rates, some of the highest in fashion, because fit and comfort are nearly impossible to judge without physically trying the item on
• Structured garments like blazers and tailored pants see outsized benefit from dimensional fit prediction, because chest, sleeve, and waist are discrete, measurable variables
• Draped fabrics, stretch materials, and layering pieces see smaller reductions, because fabric behavior under wear dominates the fit experience in ways no visualization captures

The 23 percent average is not a flat outcome. It is an average sitting across a range where some categories see 40 percent reduction and others see almost nothing. The brands seeing the higher numbers are the ones where the technology is answering a specific, measurable question: will this garment fit these dimensions. Not: does this garment look good on this kind of body.

Some of That Add-to-Cart Lift Is Not a Saved Return. It Is a Different Order.

This is the part that gets left out of most vendor conversations. Not all of the add-to-cart lift represents a prevented return. Some of it is prevented bracket buying, and some of it is just more orders.

If a shopper was uncertain and planning to bracket two sizes, and the visualization makes them confident enough to order just one, that is a real prevented return. But if a shopper’s uncertainty spans multiple size categories entirely, visualization only solves part of the problem. They may still order two sizes, expecting one to fit better, fully aware going in that one is coming back.

Nearly two thirds of consumers admit to at least one costly returns behavior, from bracketing to wardrobing to sending back different items than what was purchased. Not all of that is sizing uncertainty. Some of it is style preference, comparison shopping, or simply the fact that returns are free and easy.

The Real Reconciliation

Here is how the conversion number and the return number actually fit together. Virtual try-on increases orders. It does not reduce returns at the same rate, because some of the customers who now order without hesitation are customers who would not have ordered at all without the visualization, and a portion of them will still return.

That is not a bad outcome. It is a different outcome. More orders at a stable return rate is margin positive for most retailers. But it is not the same claim as “virtual try-on solves returns.” It solves the conviction problem. It moves shoppers from uncertain browsers to confident buyers. Some of those buyers then return for reasons the technology was never built to address.

Deloitte found retailers using AR or AI tools see a 40 percent increase in conversion and a 20 percent increase in average order value against retailers who do not. That number is real and it matters on its own. It is a conversion and AOV story first. The return story is smaller, real, but smaller, because it is competing against bracketing, subjective preference, and the gap between a rendering and a physical garment on an actual body.

What This Means If You Are Deciding Whether to Implement This

If your goal is pure return reduction, the metric to track is not add-to-cart lift. It is return rate on orders placed after try-on engagement, compared against orders placed without it. Even that comparison has a built-in skew, because shoppers who use try-on chose to use it because they were already uncertain. That makes them a higher-risk cohort from the start. A 20 percent return reduction inside that cohort is meaningful. A 40 percent reduction in a structured category like tailoring is exceptional. But expecting any visualization tool to take a 22 percent overall online return rate down to single digits assumes fit is the only thing driving returns. It is not. Style preference, color interpretation on screen, and sizing variance across brands are all doing their own damage, and none of those are fit problems a body scan can solve.

The Honest Version of the Pitch

Virtual try-on increases the odds that a customer who uses it completes the purchase. That part is proven, real, and worth building for on its own. It reduces the return rate on those orders too, just by a smaller margin than it lifts conversion, because a bigger, more confident customer pool naturally reflects a wider spread of fit needs and preferences than a smaller, more cautious one did before.

The honest answer to “does this fix conversion or does this fix returns” is both, at very different magnitudes. The conversion lift is immediate and large. The return reduction is real but smaller, because it is fighting bracketing behavior, subjective taste, and the physical gap between a rendering and a garment worn on an actual body. The ROI case for this technology should be built on conversion lift net of return impact, not on returns alone. It solves the confidence crisis at checkout. The returns crisis is a sizing variance and bracketing problem, and that one takes more than a better visualization to fix.

That is the piece we are trying to build toward at Tuck: fit prediction, not just visualization, because the data makes it pretty clear those are two different products wearing the same UI.

If you are running an apparel brand and you have been staring at a conversion lift that did not show up the same way in your return rate, I would like to hear what you are seeing. That gap is where the real work is.