3D MODELING FOR ECOMMERCE: WHAT IT TAKES TO PUT PRODUCT MODELS ON A STORE PAGE

Product photography answers one question: what does this thing look like from the angle the photographer chose. A 3D model answers every question the shopper has, because the shopper picks the angle. That is the whole difference, and it is why 3D modeling for ecommerce stopped being an experiment somewhere around the time Shopify, Amazon and Google all built native 3D viewers into their own products.

This page is the practical version of the subject. Not the visionary one. What a 3D model on a product page is made of, what your platform can and cannot display, what you need to hand a studio to get one, and what it costs. We build these models for product brands every week, so most of what follows comes from projects rather than from a whitepaper.

What 3D ecommerce actually means

The term covers four different things that get mixed up constantly, and the mix up is expensive. A brand asks for 3D, gets a 360 spin, and wonders why it cannot be used in augmented reality. Here is the honest separation.

Only the third and fourth rows are 3D ecommerce in the strict sense. The first two are output, not assets. And this matters for budgeting: an interactive model is one asset that feeds the product page, the AR button, your ads, your lifestyle stills and your catalog. A spin feeds exactly one slot on one page.

Where 3D models earn their keep in a store

A model that only sits in a viewer on the product page is an underused asset. In the projects that pay for themselves, the same file does five jobs.

• The product page viewer. The shopper rotates the item, zooms into the stitching, the port layout, the hinge. This is the job everyone thinks of first.
• The AR button. Same model, one extra format, and the shopper sees a sofa against their own wall or a watch on their own wrist. More on how that plays out in our piece on AR product visualization for ecommerce.
• Product imagery at scale. Once the model exists, every catalog shot, every colorway, every packshot is a render, not a reshoot. Twelve finishes cost about what one finish costs.
• Configurators. If the product has options, the model becomes a live customizer. We wrote up the mechanics in 3D product modeling for customization.
• Marketplace and ad creative. Amazon listings, video, social. The source is the same geometry, which is also why brand consistency stops being a fight.

What a store ready 3D model is made of

This is the part most articles skip, and it is the part that decides whether the model works on a phone over mobile data or spins the loader forever. A model for a product page is not the same object as a model for a print render. It is built to a budget.

• Geometry and polycount. A hero render can carry millions of polygons. A web model usually lives between roughly 30k and 150k triangles depending on the category. Detail that does not survive at that budget gets baked into textures instead of modeled.
• UV layout and PBR textures. The surface behavior comes from texture maps: base color, roughness, metalness, normal. Physically based rendering is what makes brushed aluminum read as brushed aluminum in any lighting the viewer applies.
• Texture resolution. 2K maps are the usual working point for web. 4K when the shopper is expected to zoom into fine grain or engraving.
• Level of detail. Some setups ship a lighter version for phones and a heavier one for desktop.
• File size. The practical target is under 5 MB for a single product, and closer to 2 MB if you care about mobile. Draco or Meshopt compression on the geometry usually gets you there without visible loss.

Then there is the format question, which is genuinely simple once someone writes it down.

So a normal delivery for a product page is a GLB plus a USDZ, plus the texture set, plus a set of rendered stills from the same scene. If a vendor quotes 3D and delivers only images, you have bought pictures.

Worth knowing that this is not vendor folklore. The Khronos Group, the same body behind glTF and Vulkan, runs a 3D Commerce working group specifically to keep product models consistent across retailers, which tells you how settled the format question now is.

Which ecommerce platforms support 3D

Short answer: all the major ones, with different amounts of effort.

• Shopify. Native. You upload a GLB as product media and the theme renders a viewer with an AR button, no app required in most themes.
• WooCommerce. Through plugins or a direct model viewer embed. Slightly more manual, fully doable.
• Adobe Commerce and Magento. Extension or custom module on the product media gallery.
• BigCommerce. Supported through apps and custom storefront work.
• Custom storefronts. Google model viewer is an open source web component. It is a script tag and a source attribute, and it handles AR routing to Quick Look on iOS and Scene Viewer on Android by itself.

The pattern here is worth noticing. The integration is not the hard part anymore. The hard part is having good models, correctly built, at catalog scale.

How the models get built

Three routes, and which one you are on depends on what you can hand over.

• From CAD. If the product was engineered, a STEP file exists somewhere. This is the fastest and most accurate route. The engineering mesh is not usable as is, it gets retopologized down to a web budget, but the dimensions are exact.
• From photos, drawings and specs. The common case for furniture, decor, packaging. The studio models by hand from references, and the accuracy depends entirely on the quality of what you send. Dimensions and material notes matter more than beautiful photos.
• Photogrammetry or scanning. Good for organic shapes and existing physical samples. Produces heavy dense meshes that still need cleanup and retopology, so it saves less time than people expect on hard surface products.

What a studio actually needs from you on day one: dimensions, a material list, photos of every side including the underside and the connections, and one real sample photo in daylight for color reference. Brands that send this get a first draft that is close. Brands that send three marketing photos get three rounds of revisions. If you are choosing tools instead of a studio, our roundup of 3D modeling software covers the paid and free options.

What actually changes in the numbers

Two effects, and they pull in different directions on your P&L, which is why 3D tends to survive budget review.

The first is returns. The National Retail Federation puts the return rate for online sales at 19.3 percent in 2025, against 849.9 billion dollars of returns across retail as a whole. A meaningful share of that is the it did not look like this problem, which is precisely the problem a rotatable model at true scale addresses.

The second is conversion. Shopify reports that on Rebecca Minkoff product pages shoppers were 27 percent more likely to place an order after interacting with a product in 3D, and 65 percent more likely to buy after engaging with AR. That is one brand in fashion accessories rather than an industry average, so treat it as an indication of the ceiling, not a promise for your category.

The honest caveat: results track how visible and how good the model is. A 3D tab buried behind four thumbnails, holding a model that reads as plastic, moves nothing. Placement and quality are not details here, they are the mechanism.

Budget and timeline

Ranges rather than a price list, because a phone case and a modular sofa are not the same job.

Catalog scale changes the math in your favor. A hundred products in the same family share materials, lighting and setup, so the per unit cost drops well below the one off price. If you want the numbers for your specific catalog, our 3D product modeling services page is the place to start, and 3D product rendering covers the imagery side.

How to start without betting the whole catalog

Nobody should model 4,000 SKUs to find out whether this works for them. The sequence that works.

• Pick five to ten products. Best sellers, or the ones with the worst return rate. Both give you a clean read.
• Do them properly. A cheap pilot that looks cheap will kill the initiative for two years.
• Put the viewer where it is seen. First media slot, visible 3D and AR affordance, not a tab.
• Measure the right things. Viewer engagement rate, add to cart on those pages against a control group, and return rate over a full return window. The last one takes a quarter, so start the clock early.
• Then scale by family. Reuse materials across the range.

One more thing that gets forgotten: the model is an asset you own. Photography is a service you rebuy every time the product changes color, angle or context. That is the quieter argument for 3D, and over a couple of seasons it is usually the decisive one.

FAQ

If you have a catalog and a shortlist of products worth testing, tell us the category and what source files exist. We will come back with what is realistic to build, in what order, and what it takes.