What is deployed today
Published XR retail work divides cleanly into campaign activations, which are short-lived and reach-driven, and catalogue tools, which are maintained and conversion-driven.
The two are commissioned by different people, measured on different numbers and priced differently, yet they are frequently discussed as one category. A brand team buys an activation to be noticed. An ecommerce team buys a try-on or a configurator to reduce a specific doubt before checkout. Confusing them is the fastest way to spend an activation budget on something that needed to be maintained for three years.
Activation work is the larger group in the published record. Aircards publishes browser-reached activations for Nike and Foot Locker. FFFACE.ME documents in-store work for Bershka and a Clarins filter. These run for a season, are reached without an install and are measured on engagement and reach.
Catalogue work is smaller in the record but more durable. Try-on for eyewear, cosmetics and footwear, and placement for furniture, address a specific pre-purchase doubt and are maintained alongside the product range. The value compounds with catalogue size, which is why the economics differ so sharply from a campaign.
A third group covers the store itself rather than the product, including navigation, in-store displays and staff training. It appears in the record but is thinner, and it is usually commissioned by operations rather than by marketing.
What constrains it
The binding constraint in retail XR is the asset pipeline, not the application, and it scales with catalogue size rather than with feature count.
A try-on experience for one product is a project. The same experience across four thousand SKUs is a production system, and the difference is not in the software. Every product needs a three-dimensional asset that is accurate enough to inform a purchase and light enough to load on a mid-range phone over mobile data. Product photography does not convert into that automatically, and manufacturing CAD, where it exists, carries far more detail than a device can draw.
Two further constraints follow from the channel rather than the technology.
Friction decides whether anything gets used. A shopper deciding between two products will not install an application, which is why so much of this work is delivered through the browser and why asset budgets are tight. An experience that is technically excellent and reached through an install will show good session metrics from a tiny audience.
Catalogue churn sets the recurring cost. Fashion ranges turn over seasonally, and an asset pipeline that cannot keep pace produces an experience covering last season. This is an operating commitment rather than a project, and it is the line most often missing from a first budget.
What evidence of return exists
Most published return figures in retail XR are vendor-supplied, and very few are accompanied by a method a buyer could check.
This deserves stating directly, because the category is unusually full of confident percentages. Conversion uplift, return-rate reduction and engagement multiples circulate widely, and they are typically published by a platform or a studio describing its own work. That does not make them false. It does mean they are marketing claims rather than measurements, and they are frequently quoted without the baseline, the sample or the period.
The register records what a source states rather than assessing it, and the methodology sets out that limit. Across the case studies currently on record, descriptions of what was delivered are common and published outcome figures with a stated method are rare, which matches the pattern across the industry.
Two questions make any quoted figure useful or expose it. Against what baseline, and over what period. A conversion uplift measured against no comparison, or a return-rate reduction over a single season on a single product line, may be genuine and is not evidence that the same will happen elsewhere.
The practical position is to treat published figures as a reason to run a test rather than as a substitute for one. Retail is well instrumented, and a bounded pilot on a subset of the catalogue produces a number that belongs to the buyer and answers the question for their products.
What a studio must prove
Ask what a hundred additional products would cost and how long they would take, because that answer reveals whether a pipeline exists.
A studio with production capability answers with a per-product figure and a throughput. A studio without one quotes another project, because each product is being handled bespoke. For catalogue work that difference determines whether the engagement scales or stalls at the pilot.
Three further checks are worth the time in this sector.
Ask how the experience is reached. A studio experienced in retail will discuss browser delivery, load time and the drop-off between a scan and a loaded experience without prompting, because those numbers decide whether an activation performs.
Ask who maintains assets as the range changes, and what that costs annually. The answer separates a studio selling a project from one selling an operating capability, and both are legitimate provided the buyer knows which is on offer.
Ask for delivered work with the same catalogue shape, not the same industry. Experience across a handful of hero products transfers poorly to thousands of variants, and the pipeline question above is where that shows.
Where projects fail in this sector
Retail XR projects fail on asset maintenance and on unclear success measures, more often than on the build.
Maintenance is the common one. An activation ends cleanly, but a catalogue tool that is not maintained degrades quietly: products disappear from the range, assets go stale and the experience quietly stops matching what is for sale. Ownership of that work belongs in the original agreement rather than in a renewal conversation.
Unclear measurement is the expensive one. A project commissioned by marketing for reach and judged later by ecommerce on conversion will fail that second test regardless of quality, because it was not built for it. Agreeing the measure before the build is the cheapest risk reduction available, and it also determines what instrumentation the studio needs to include.
The third is scope drift from activation into catalogue. A successful seasonal activation invites a request to extend it across the range, and the underlying build is frequently not structured for that. Deciding at the outset whether the work is disposable or foundational avoids paying twice.
What the published record does not tell you
Retail case studies report reach and rarely report conversion.
About one in sixteen of the retail work here names a platform and the same proportion carries a date. Where a figure does appear it is usually an engagement measure such as sessions, dwell time or shares, and those are the numbers a campaign is briefed on. What almost never appears is whether the activation moved a commercial outcome, because that figure belongs to the retailer and is rarely released.
This produces a specific risk at shortlisting. A portfolio full of recognizable brands demonstrates that a studio can deliver to a brand's standards and under a campaign deadline, which is genuinely difficult, and demonstrates nothing about returns. Those are separate claims and a case study usually supports only the first.
The second unstated variable is how the experience was reached. Web-based activations that open from a code or a link reach anyone with a phone and are constrained by browser capability, while application-based work reaches a smaller and more committed audience with fewer technical limits. A case study describing a campaign without naming the delivery route has left out the thing that determined its ceiling.
Ask what the retailer measured, whether the studio saw it, and how the experience was distributed. A studio comfortable answering all three has usually worked with clients who tracked the result.
Further published work in this category
Retail work in the register splits between in-store deployment, campaign activation and content production, and the three are commissioned by different teams.
Draw and Code publishes in-store Augmented Reality built for John Lewis and Partners, which is a deployment inside a physical estate and therefore involves store operations rather than marketing alone. Next/Now publishes a product showcase for Weber Grills, which is the same technique aimed at explaining a large product a shopper cannot easily assess in a shop.
Two further patterns are worth separating from try-on work. FFFACE.ME publishes computer generated imagery produced for a Shiseido product launch, which is content production rather than an interactive build and is commissioned on a content budget. Program-Ace publishes an active threat simulation delivered for Gap Inc., which is retail as an employer rather than retail as a storefront, and it is a reminder that the sector commissions workforce training as readily as customer experience.
How deep the field actually is
Retail is among the most commonly claimed sectors in the register and among the thinner ones on published evidence.
Under half the studios on record publish retail work, and the ones that do are concentrated in campaign and activation rather than spread across the sector. Try-on, out-of-home and brand experience dominate. In-store operational deployment, catalogue-scale visualization and post-purchase support appear far less often, even though those are the areas a retailer with an existing digital estate is most likely to be procuring for.
That imbalance should change how a shortlist is read. A studio with a strong campaign portfolio has demonstrated speed, brand handling and delivery under a fixed date. It has not thereby demonstrated that it can operate inside a retail technology stack, integrate with a product catalogue or support a deployment across an estate for several years, and those are different capabilities with different evidence.
Ask which retail projects are still running. Campaign work is designed to end, so a portfolio of completed activations is entirely normal and says nothing either way. A studio that can name something still live in a store, and describe who maintains it, is answering the question a longer deployment actually turns on.
Two further things separate a campaign supplier from a deployment partner, and both are visible before signing. The first is who holds the relationship with the retailer's technology function. Campaign work is usually commissioned by marketing and never touches store systems, so a studio may have delivered for a retailer repeatedly without ever meeting the people who would have to approve an in-store installation. The second is what happens to the experience when the campaign ends. An activation built to be switched off is straightforward, while anything intended to persist needs an owner, a support route and a plan for the day the underlying platform changes.
Neither point argues against campaign studios. It argues for matching the studio to the job, and for being explicit about which of the two jobs is being bought, because the portfolio that proves one says very little about the other.
The measurement gap is worth one closing note, because it is the thing most likely to decide whether a second project is funded. A retailer that defines the success measure before the brief goes out retains the ability to judge the result, while one that accepts engagement figures afterwards has agreed to be told what happened by the party that delivered it. Neither outcome reflects on the studio. The difference is set on the buyer's side, and it is set early.
Where to go next
Every studio named here is on the public record with its sources and its verification date, and the full list sits on the register.
Rankings are scoped rather than global, because a studio strong in seasonal activation is not thereby strong in catalogue production, and those are the two distinct capabilities this sector buys. Retail work in the register is dominated by one technology and one use case. Most studios with published retail work also appear under Augmented Reality development, which is the technology behind almost all published try-on and in-store work, and the job those builds do is documented in more depth on the product visualization pillar.
