Virtual try-on technology for ecommerce: A retailer's guide

Does virtual try-on actually reduce returns, or does it just delay them?

Most retailers assume it's the first, since showing a shopper what something looks like before they buy should mean they order the wrong thing less often. That only holds if the preview is telling the truth, and usually the render is the problem, smoothing over how a garment actually fits, or making a colour look better under screen light than it ever will in daylight. Either way, the shopper doesn't find out until the parcel's already arrived.

Those two failures aren't hypothetical, and they're also the two biggest reasons UK shoppers send things back. Royal Mail's small business guidance, drawing on ZigZag and Retail Economics figures, puts online returns at £27bn in 2024, and 2025 YouGov research cited in the same guidance splits the blame almost evenly, 46% down to fit and 46% down to the product looking different once it arrived, whether that's colour, quality or size.

How much virtual try-on actually helps comes down to whether it solves those two problems for real.

This guide covers what virtual try-on actually is, how it works, which categories benefit most, what the evidence says about returns and conversion, and how to implement and measure it properly.

What is virtual try-on?

Virtual try-on technology lets shoppers see how a product might look before they buy it. They can preview an item on themselves using a live camera, apply it to an uploaded photo, or view it on a representative model.

Augmented reality places the product onto the live camera feed or image, and computer vision works out exactly where it should sit and how it should align with the body or face. AI then generates or refines the result into a realistic preview, drawing on a digital representation of the product behind it, a 2D image, a 3D model, or detailed measurements.

It can look similar to a social media filter that places sunglasses or a hat on someone's face. This experience is tied to a retailer's actual product catalogue, linking it to real products, colours, variants and SKUs that shoppers can purchase immediately.

It's designed to show shoppers how a product looks on them, not whether it will fit, though many retailers market it as a solution for reducing returns anyway. Unless the system also uses body measurements and product sizing data, it cannot tell shoppers whether a garment will actually fit.

The term itself is also used as a catch-all for a few related technologies.

How does virtual try-on work?

From the shopper's perspective, virtual try-on feels almost instant. Open the camera or upload a photo, choose a product, and the preview appears almost immediately. That simple experience still requires the software to analyse the shopper, understand the product, and bring the two together for the final preview.

1. Starts with the shopper

Virtual try-on requires a visual reference to start with. A shopper might use a live camera, upload a photo, select a model close to their own body type, or enter their own measurements directly.

The input depends on the product. Someone browsing glasses or lipstick usually wants to compare a few options, so a live camera works well there, whereas trying on outfits calls for a full-length photo, since it gives a much clearer sense of the shopper's actual proportions and pose.

2. The software works out where the product belongs

The first step is identifying the right body feature. For eyewear, the software needs to map the eyes, nose, and outline of the face. Jewellery would look for a finger, wrist, ear, or neck. Clothing is more challenging, since it needs to read the position and shape of the entire body rather than one fixed point.

In a live try-on, these reference points are tracked continuously, so the product stays in place as the shopper turns their head or moves their hand.

3. A digital version of the product

The product needs its own digital version too, and how much data that takes scales with the category. Eyewear needs accurate frame dimensions, since an overlay only looks right if the glasses render at their true width against the shopper's face. Jewellery needs the same kind of sizing data, a ring has to be scaled to an actual diameter, not just however large it happens to look in a photo. Clothing needs by far the most actual garment measurements like shoulder width, sleeve length and how the fabric is cut, compared against the shopper's own body.

In practice, most systems don't get that far. A 2026 University of Washington study, presented at SIGGRAPH, traced this to a gap in the training data. The image sets used to build these tools are made up almost entirely of garments that already fit the person wearing them, with very few examples of a bad fit. Take a shopper who wears a medium trying on a dress only available in extra-small, a dress that would visibly pull at the shoulders or gape at the waist in real life. The tool shows the dress fitting well anyway, since it was mostly trained on garments that already fit.

4. The preview is created using one of two technologies: augmented reality or generative AI

Live AR uses the phone's camera to place a digital version of the product onto the shopper in real time. As the shopper moves, the product moves with them, the same way Pandora's jewellery try-on works, covered in more detail in the category breakdown below.

Generative AI skips the live camera altogether. Shoppers upload a single photo and select a product, and the AI generates a new image showing how the item would look on them. Goddiva introduced this experience in January 2026 using Google Gemini alongside its own AI models, generating previews based on the shopper's body shape and proportions.

A few retailers let shoppers use a model or avatar instead of their own photo. That only works well for comparing styles, since a model can't reflect a shopper's actual body.

In-store virtual mirrors move the same experience onto the shop floor, useful for staff helping customers. But they cost more to run than a digital tool, since the retailer needs real hardware, a physical space for it, and staff trained to keep it maintained and running.

5. It brings the shopper back into the buying journey

This step ties the preview to an actual sale. The system needs to track the exact SKU, colour and variant the shopper is browsing, so the same product gets added to their basket. Shoppers should also be able to move easily between products, comparing colours or styles without losing their place on the page. Retailers, with consent, can measure which products are tried on most often and how those interactions influence purchases.

Which retail categories benefit most?

A shopper choosing between two shades of lipstick has a very different problem to someone wondering whether a jacket will actually fit. Virtual try-on helps with both, though how much it helps depends entirely on what's being sold.

1. Eyewear

A catalogue photo shows what a frame looks like, but it cannot show how a specific width, bridge shape or colour will suit an individual face. The only way to answer that is to see the frame on an actual face rather than a stock model. Specsavers built Frame Styler in 2018 to do exactly that, letting shoppers preview a frame on their own face before buying. 

2. Beauty and cosmetics

A lipstick shade can look very different on a person's face than it does on a product page. Screen colour and skin tone both affect how a shade actually reads, and no amount of product photography fully solves that. Charlotte Tilbury built Magic Mirror with Holition in 2016 to let shoppers see a shade applied to their own face before buying.

3. Jewellery and watches

Rings and bracelets are often gifts and significant purchases, so customers want a real sense of scale before committing. A product photo gives little sense of that, since there's nothing in the image to compare it against. Pandora launched its live AR virtual try-on with Tangiblee in 2020, allowing shoppers to see a ring at its actual size on their own hand.

4. Fashion and apparel

Fit is one of the two biggest reasons UK shoppers return things, as covered earlier in this guide, and clothing is where that problem shows up most directly. A product photo shows how a garment looks in isolation, not how it will actually sit on a specific body. Zara launched an AI-powered virtual fitting room in its app in December 2025, building a 3D avatar from two photos, a headshot and a full-body shot, that walks and turns so shoppers see the fit from multiple angles instead of a single flat image.

5. Footwear and accessories

Shoe size depends on foot width, shape and volume as much as length, and none of that comes through in a single photo. Nike built Nike Fit around this problem in 2019, using a smartphone camera to capture 13 data points, accurate to within 2 millimetres, and matching them against the internal dimensions of each shoe.

6. Furniture and other categories

Furniture and other large items use AR product visualisation instead, a distinction already drawn earlier in this guide, placing products into a shopper's own room instead of onto a body. The technology is built for a different situation, though it's still answering something similar, what an item will actually look like in the space where it'll be used.

How to implement virtual try-on in an ecommerce store?

Step 1: Define the customer problem

The starting point is understanding why virtual try-on is being introduced in the first place. For some retailers, the goal is reducing returns caused by sizing or fit issues. For others, it's helping customers choose the right colour, see how a product looks on them, or simply make product pages more engaging. Getting this clear early makes the rest of the decisions, which method, how to measure it, much easier to make.

Step 2: Choose the right try-on method

Products need different types of virtual try-on depending on what's being sold. Live AR works well for glasses, jewellery and makeup, where customers want to see the product instantly. Photo-based AI suits clothing better, since shoppers want to see an outfit on their own body, not just on a model. AI avatars offer a privacy-friendly alternative for shoppers who'd rather not upload a photo, while smart mirrors and in-store kiosks extend the same experience into physical retail environments.

Step 3: Prepare product data

Virtual try-on depends on accurate product data. Each product needs high-quality images, correct dimensions, colour and material information, and 2D or 3D assets where the category calls for them. Every asset also needs to be linked to the correct SKU and variant, so customers see exactly what they're buying and not a mismatch that only shows up once the order arrives.

Step 4: Integrate with the ecommerce platform

The integration approach depends on the platform already in place. Retailers using platforms like Shopify often start with a plugin, since it's quick to deploy. Businesses running custom websites or apps usually integrate through an SDK instead, and larger operations with more complex catalogues sometimes build directly on a provider's API. Those with physical stores also need to decide whether the same solution should extend to smart mirrors or kiosks on the shop floor.

Step 5: Design the shopping experience

The "try on" button should be placed strategically, somewhere a shopper won't need to search for it, and a brief note nearby should explain why camera or photo access is needed before that request appears. Inside the experience itself, customers should be able to switch between colours and variants without leaving the try-on view, and product details along with the add-to-basket button need to stay visible throughout. Standard product images should stay available too, for anyone who'd rather not use the feature at all.

Step 6: Plan for privacy

Any feature using a customer's photos or live camera feed needs privacy built into the plan from the start. That means deciding what data gets collected, where it's processed, how long it's stored, and who's responsible for managing it. Customers need clear information before uploading a photo or enabling their camera, and opting in should be genuinely easy. Reviewing the relevant legal and regulatory requirements before launch saves a much harder conversation after it.

Step 7: Test before launch

Testing needs to run across different devices, browsers and network connections before anything goes live. Lighting conditions, skin tones and body types all affect how well the experience performs, alongside loading time, image quality and tracking accuracy. The product page itself should keep working smoothly for anyone who chooses not to use virtual try-on, and accessibility deserves the same scrutiny as everything else on this list.

Step 8: Start small and expand

There's no need to launch a virtual try-on across an entire catalogue at once. A smaller group of products, chosen because customers are most likely to benefit from the feature, makes a far more useful starting point. Tracking engagement, conversion, customer feedback and return reasons during that period shows what's actually working before the rollout goes any further.

How should retailers measure virtual try-on ROI?

1. Adoption and usability

This starts with basic usage. How many product-page visitors actually open the try-on feature, and how many agree to camera access where it's needed. From there, it's worth checking whether sessions complete properly, how long it takes to reach a first preview, and how often shoppers hit an error or give up partway through.

2. Engagement

Once shoppers are using the feature, it's worth looking at how they use it. Are they trying several products in one session, or switching between variants to compare colours and styles? Repeat visits to try-on, more time spent on the product page, and shoppers saving or sharing what they've tried all suggest genuine interest is building.

3. Commercial performance

Add-to-basket rate and conversion rate are the two most direct numbers here, whether virtual try-on actually gets a shopper closer to buying. Average order value and revenue per visitor matter too, since spend can go up even when conversion doesn't move much, and it's worth tracking multi-product purchases as well, since someone comparing a few items in one session may end up buying more than one.

4. Post-purchase performance

Splitting returns by reason separates genuine fit problems from colour, quality or expectation issues that virtual try-on was never going to fix. Comparing the return rate among try-on users against those who didn't use it gives the clearest sense of whether the feature is actually changing anything. Refund costs, reverse logistics and customer satisfaction fill out the rest of the picture.

5. Measuring it properly

Shoppers who open virtual try-on are usually more engaged from the start, sometimes already leaning toward a purchase, so comparing that group directly against everyone else makes the feature look better than it actually is. A proper A/B test, or matched groups where only one side gets access to virtual try-on, gives a much more honest read on what it's actually doing to sales, returns and behaviour.

What should retailers look for in a virtual try-on platform?

1. Category-specific accuracy: A platform built for eyewear won't necessarily perform well for apparel, footwear, or cosmetics. Look for proven accuracy in the specific category being sold.

2. Fit prediction: Ask whether the platform uses actual body and garment measurements to predict fit, or simply generates a realistic-looking image.

3. Product onboarding: Understand how products are digitised and catalogued. Some vendors handle most of the setup, while others expect retailers to manage much of the process.

4. Integration flexibility: The platform should work across web, mobile apps, and in-store experiences, with SDKs, APIs, or plugins that fit the retailer's existing technology stack.

5. Mobile performance: Since most customers will use virtual try-on on their phones, the experience should be fast, responsive, and reliable on mobile devices.

6. Built-in analytics: Look for reporting on adoption, engagement, and conversion, so business impact can be measured without stitching together data from multiple tools.

7. Privacy, security, and compliance: Review how customer data is collected, stored, and processed. Security certifications and clear data-handling policies should go in front of legal and procurement before any contract is signed.

8. Inclusive testing: Ensure the platform supports a range of body types, skin tones, devices, and lighting conditions to deliver consistent results.

9. Ongoing model improvements: Ask how the vendor maintains and improves accuracy after deployment. The best platforms keep learning and refining over time, while others stay fixed at whatever version was first delivered.

10. Total cost of ownership: Consider implementation, integration, maintenance, and support costs alongside licensing fees to understand the platform's true long-term investment.

Bringing virtual try-on to your store with Fynd GlamAR

Fynd GlamAR brings virtual try-on to eyewear, makeup, jewellery, watches and accessories, so shoppers can see how something actually looks on them before they buy it, whether they're on their phone, on a laptop, or standing in front of a screen in-store.

GlamAR manages everything from 3D asset creation and catalogue onboarding to integration through an SDK, API or plugin, with most retailers able to go live using a short code snippet or, on Shopify, a dedicated app. Beyond implementation, the platform provides detailed analytics into shopper engagement and product interactions.

Low Cost Glasses sells eyewear entirely online, with no physical stores to fall back on. Its 2D try-on couldn't show subtle mismatches in frame size, colour or alignment, so shoppers who noticed something felt off tended to hesitate and drop off before buying. After switching to GlamAR's 3D virtual try-on, letting shoppers rotate a frame and check it from multiple angles, monthly orders rose from 350 to more than 500, a 35% increase.

See what virtual try-on could do for your catalogue.

Explore Fynd GlamAR