The future of 3D in e-commerce: what becomes standard, and what does not

Predictions about 3D and e-commerce usually describe a world that is technically possible and commercially unlikely: metaverse showrooms, avatars with your exact body, headsets on every sofa. A more useful exercise is to look at what phones and browsers already do reliably, work out what that makes cheap, and assume the cheap things win. On that basis, here is what we think a mainstream product page carries in three years, and what it does not.

The argument from devices

Two platform facts do most of the work here.

On Android, Scene Viewer puts a glTF or GLB model into the camera view from an ordinary web link, on any ARCore device running Android 7.0 or later, with no app install. On iPhone and iPad, AR Quick Look does the same with a USDZ file from Safari, on iOS 12 and later.

That is the whole story of why AR stopped being a gimmick. The cost of showing a product in someone's room fell from "build an app and persuade them to install it" to "host two files and add a button". Anything that gets that much cheaper tends to become normal.

What becomes standard

A rotatable 3D model on the product page. This is already happening in furniture, lighting, appliances and eyewear. It is cheap, it works on every device including desktop, and it does not require the shopper to understand anything new. In three years its absence on a considered purchase will look like a missing zoom function looks today.

An AR button next to it. Not because everyone uses it, but because it costs almost nothing once the model exists, and for the minority who do use it the effect on confidence is large. The engineering is a single attribute on a `` element.

Real dimensions attached to the model. The spec table stays, but the measurements move onto the object where the shopper is already looking.

Configurators for anything with variants. If a sofa comes in forty fabrics, photographing forty sofas is not a strategy. MELIMELI and Sweef both moved to configurators for exactly this reason, and the economics only get more one-sided as catalogues grow.

What does not become standard

Headset shopping. Vision Pro and its competitors are real products with real users, and those users are not a retail channel. Building a storefront for a headset audience in 2029 will still be a marketing exercise rather than a revenue line, unless something changes in the install base by an order of magnitude.

Full-body avatars for garment fit. Fabric simulation on a body whose measurements the phone does not actually know is the hardest problem in this field. It will keep improving and it will keep being a styling preview rather than a fitting room. Expect progress in accessories, where the object is rigid, long before trousers.

The metaverse store. A 3D shop you walk around in performs worse than a grid of products on nearly every metric a retailer cares about. Every few years someone rebuilds it. It does not stick because the product page is not the part of shopping that needed to be more immersive.

Scanning replacing modelling. Phone-based scanning is genuinely useful for rough geometry and getting worse-than-studio textures fast. For a product page, where the model has to look like the thing you sell under controlled lighting, an artist still cleans it up. Scanning shortens the job rather than removing it.

The part that decides adoption

The bottleneck is not the viewer. It is the asset pipeline. A retailer with 3,000 SKUs is not blocked on whether AR works, they are blocked on who makes 3,000 models, keeps them consistent, updates them when the product changes and hosts them at a size that loads on mobile.

That is why the commercial shape of this industry is moving from projects to subscriptions. You do not buy a model once, you keep a catalogue current. Our own pricing is built that way on purpose: Basic from 10 dollars per product per month on a twelve-month term with a minimum of 100 variants, Value from 20 dollars per model per month on twenty-four months, and a custom Enterprise tier. Whether you use us or someone else, the retailers who get this right in three years will be the ones who treated 3D as a maintained catalogue rather than a campaign.

A reasonable timeline

In the next year, expect 3D and AR to keep spreading in the categories where scale and material matter: furniture, lighting, appliances, heating, eyewear. In two, expect it in adjacent hard goods, sports equipment, garden products, bathroom fittings. In three, expect it to be table stakes in those categories and still patchy everywhere else, because nobody needs to see a phone charger in their room.

The prediction we are most confident about is unglamorous: the winning implementations will be the fast, quiet ones. A model that loads in under two seconds, looks like the product, and has a button that works on the shopper's actual phone. Everything else on the futurist slide deck is optional.

Where to start if you believe this

Do not start with a catalogue rollout. Start with one product, measure it, and let the number decide the budget. We build the first model free for exactly that reason. Send us a product, or look at what we build if you already know the shape of the project.

Sources

• Google, Scene Viewer developer documentation.
• Apple, AR Quick Look.
• Shopify, 3D in e-commerce: merchant results and model requirements.