- Why Virtual Try-On Vendors Look Identical on a Demo Call
- Eight Criteria for Choosing a Virtual Try-On Vendor
- How to Run a Virtual Try-On Pilot Test That Proves Something
- The Virtual Try-On RFP Questions That Get Straight Answers
- Frequently Asked Questions
- Conclusion
Every vendor demo ends with a garment rendered on a body, and they all look convincing. The direct answer on how to choose a virtual try-on vendor: the render is the commodity, and the decision is settled by what sits around it, the size call, the pricing unit, the catalogue coverage and the measurement plan. This post sets out the criteria worth scoring, how to design a pilot, and what to put in writing before anyone signs.
Why Virtual Try-On Vendors Look Identical on a Demo Call
A demo runs on a vendor-chosen product, photo and body type. Under those conditions every tool in the category performs well, which is exactly why a demo ranks vendors badly.
What is being bought is not visualisation, it is certainty before checkout. The European Environment Agency puts the average EU return rate for clothing at 20%, and for footwear at 30%. Research by fulfilmentcrowd, reported by just-style in 2025, found 30% of fashion items bought online in the UK come back each year, with poor fit the leading reason.
A virtual try-on render answers how a piece looks, not whether to take the M or the L, which is where most of that volume originates. A vendor that stops there leaves the expensive question open, and that gap is invisible on a demo.
Eight Criteria for Choosing a Virtual Try-On Vendor
Score every shortlisted vendor on the same eight lines. The first three decide whether the tool moves your return rate at all, the rest whether the rollout survives contact with your catalogue.
- Catalogue coverage. Ask which product types render reliably and which do not. Knitwear, tailoring, prints and accessories behave differently, and a tool that handles only flat-lay tops covers a fraction of your GMV.
- Size output, not just the image. A recommendation built from your own size guide beats a generic scale, because your M is not the market's M. The mechanics are covered in AI Size Recommendation: Fixing Fashion's Biggest Return Driver, and the capability itself in smart sizing.
- Colour advice. Colour is the second information gap on a product page. A vendor that reads the shopper's palette and says which shades suit them closes it, as the colour profile does.
- Integration cost. A script tag or a native app listing is a two-week job. A 3D asset pipeline or a body-scan requirement is a programme, with a budget and a project manager attached.
- Latency and failure behaviour. Ask what the shopper sees when a render is slow or fails. A spinner that never resolves costs more than the feature earns.
- The pricing unit. Per try-on, per session, per MAU and per SKU are not comparable. Klooset bills per try-on, €199 a month for 1,000 on Starter, which makes the cost per resolved doubt easy to compute.
- Data handling. Where the photo is processed, for how long, and by which sub-processor. This one has its own checklist in Virtual Try-On and GDPR: A Compliance Guide for Retailers, and it belongs in the contract rather than on a marketing page.
- Measurement support. The vendor should hand you an experiment design and the events to track, not a dashboard of render counts. If they cannot describe how you will attribute a return-rate change, they have not run the test before.
The five tools retailers ask us about are scored on these lines, side by side, on the comparison pages.
How to Run a Virtual Try-On Pilot Test That Proves Something
Most pilots fail on design, not on technology. One that runs across the whole catalogue with no control group produces a number nobody in the business believes.
A defensible structure: the top 20% of the catalogue by GMV, a holdout group on the standard product page, and 4 to 6 weeks. That window gives most mid-size catalogues enough return cycles to separate signal from seasonality.
Instrument three metrics before launch, not after: add-to-cart rate on the covered products, size-related return rate isolated from other reason codes, and average order value. The reason-code split is the one most teams skip, and without it a drop in returns cannot be told apart from a quiet month.
Set the bar in advance. Independent research gives a reference point: AI-powered outfitting and bundling modules produced a net ROI of over 6x across three years, a 15% lift in conversion and a 10% increase in average order value, per a Forrester Total Economic Impact study commissioned by Stylitics. Agreeing the threshold first stops the pilot being re-interpreted afterwards.
The Virtual Try-On RFP Questions That Get Straight Answers
Eight questions, in writing, before signature. Each is phrased so a vague answer is visible as one.
- Which product categories in our catalogue will not render acceptably, and what is your pass rate on them?
- Does the size recommendation use our own size guide, or a generic scale? What happens when a product has no grid?
- What are the median and 95th percentile render times on mobile, and what does the shopper see when one fails?
- What exactly counts as one billable unit, and what happens on overage?
- Where is the photo processed, how long is it retained, and which sub-processors touch it?
- Which events do you emit into our analytics, and can we attribute a return to a session that used the tool?
- Name a retailer of our size and category that has run this for more than six months.
"22-43%, or on average one third of all returned clothing bought online, ends up being destroyed." Source: European Environment Agency, 2024.
If you are scoring vendors against these criteria, book a 30-minute demo: we answer all eight questions on your own catalogue, and the ROI calculator models the payback before you commit to a pilot.
Frequently Asked Questions
What is the single most important criterion when choosing a virtual try-on vendor?
Whether the tool also resolves the size question. Poor fit is the leading reason fashion items come back, so a vendor that renders the garment without calling a size leaves the largest driver of your return rate untouched.
How long should a virtual try-on pilot run?
4 to 6 weeks on the top 20% of the catalogue by GMV, with a holdout group on the standard product page. Shorter than that and the return cycle has not completed for the orders in the test.
How do you compare virtual try-on vendors with different pricing models?
Convert every quote to a cost per completed try-on at your expected monthly volume, then compare it to your average order value. Per-MAU and per-SKU pricing can look cheaper in the proposal and land higher once real session volume arrives, which is why the business case belongs on your own numbers.
Conclusion
Choosing a virtual try-on vendor is not a visual judgement, it is a scoring exercise on catalogue coverage, size output, integration cost and measurement. Vendors that look alike on a demo separate quickly once those eight lines are filled in. For a scored answer on your own catalogue, request a pilot and we will run the criteria against the products that carry your GMV.
