Commerce·April 2025·9 min read

How to Leverage the Metaverse for Your E-commerce Brand

The metaverse-as-storefront pitch from 2021 and 2022 did not survive contact with actual commerce. What did survive, and quietly became standard, is 3D and AR product viewing built on glTF and WebXR: a shopper points a phone at their living room and sees the couch at scale, no headset and no virtual world required.

What the pitch actually was

The original claim was specific: build a branded space on Decentraland, Roblox, or Spatial.io, sell wearables as NFTs, accept crypto at checkout, and host live shopping events inside a persistent virtual world. Brands that moved fast on this between 2021 and 2022 got press coverage and not much else.

None of the mechanics generalized to ordinary retail. A virtual storefront required a customer to already be inside a specific platform, using a specific client, often on a desktop, which is not where retail traffic lives. Checkout inside these worlds routed around a brand's existing commerce stack instead of extending it.

Why it looked plausible at the time

Three things converged in 2021: Meta's rebrand made "metaverse" a board-level topic overnight, NFT trading volume was at a peak, and pandemic-era online retail growth made any new digital channel worth a press release. None of the three was evidence that virtual-world storefronts converted browsers into buyers.

A platform requiring its own client, its own avatar, and its own login before a customer can see a product is adding friction, not removing it. Retail conversion has never rewarded extra steps between interest and purchase.

What shipped instead: 3D and AR product viewing

The part of the pitch that survived is unglamorous: put a 3D model of the product on the existing product page, and let it open in AR on the customer's own phone, in their own browser, with no app install.

Two format decisions made this durable. glTF became the interchange format for the model itself, standardized by Khronos and later ratified as ISO/IEC 12113 (Khronos — glTF 2.0 Specification). The WebXR Device API gave browsers a standard way to request camera-tracked AR sessions without a proprietary SDK (W3C — WebXR Device API).

The two AR viewers that actually ship

On iOS, Apple's AR Quick Look renders a USDZ file directly from Safari with no app required, triggered by a link tag in the page (Apple Developer — Previewing a Model with AR Quick Look). On Android and desktop, Google's <model-viewer> web component wraps WebXR and Scene Viewer behind one custom element that degrades gracefully to a rotatable 3D view where AR isn't available (modelviewer.dev — the model-viewer web component, Google for Developers — AR with model-viewer).

<model-viewer
  src="chair.glb"
  ios-src="chair.usdz"
  alt="A 3D model of an armchair"
  ar
  ar-modes="webxr scene-viewer quick-look"
  camera-controls
></model-viewer>

That one element, two files (a .glb for Android and desktop, a .usdz for iOS), covers both major AR entry points without a platform SDK or a wallet connection. Shopify's product media pipeline now treats a 3D model as a native media type in its Storefront and Admin APIs, alongside images and video (Shopify.dev — Model3d object, Storefront API, Shopify.dev — Manage media for products and collections).

What did not survive, in the platforms' own numbers

Meta's own 10-K filing for fiscal 2024 states that Reality Labs investments reduced the company's overall 2024 operating profit by approximately $17.73 billion, and that losses were expected to keep increasing into 2025 (SEC EDGAR — Meta Platforms, Inc. Form 10-K, FY2024). The same filing lists Reality Labs revenue as $2.146 billion for 2024, against $164.5 billion for the company as a whole.

The more telling detail is what that revenue is attributed to: "consumer hardware products, such as Meta Quest and Ray-Ban Meta AI glasses, and related software and content." Not virtual real estate, not branded storefronts, not NFT wearables. The segment's own commercial story moved to wearable hardware.

The company that coined the term now attributes its Reality Labs revenue to headsets and AI glasses, not virtual-world commerce. That is the clearest first-party signal available that persistent branded storefronts inside a metaverse platform never became a retail channel.

Storefront pitch versus what actually shipped

2021-2022 pitch	What shipped by 2026	Where it lives now
Branded store inside Decentraland, Roblox, or Spatial.io	3D model of the product on the existing product page	Standard product media, alongside images and video
NFT-based loyalty and digital wearables	Ordinary loyalty points and account-based perks	Existing CRM and loyalty platforms
Crypto checkout inside a virtual world	Card and wallet payments through the existing gateway	Unchanged commerce stack
Live shopping events in VR platforms	Live shopping over ordinary video, on the storefront or social apps	Livestream commerce features on existing platforms
AI shopping assistant inside a 3D avatar space	Text or voice assistant on the storefront itself	Storefront search and support widgets

The mechanics browsers actually standardized on

WebXR exposes two distinct session types, and confusing them is a common source of broken demos. An immersive-vr session takes over the full display for a headset; an immersive-ar session composites 3D content over the device's live camera feed, which is the mode product viewing actually uses (W3C — WebXR Augmented Reality Module, Level 1).

Placement in the real world depends on the WebXR Hit Test API, which casts a ray from the screen into the camera's understanding of real-world surfaces and returns where a model should sit (MDN — XRHitTestSource). None of this is exposed directly to a shopper; it is what makes "tap to place" behave correctly on a real floor instead of floating mid-air.

Asset budgets nobody mentioned in 2021

A glTF file with an unoptimized mesh and full-resolution textures can run past 30 MB per product, which will show up as a Largest Contentful Paint regression on the exact page a shopper is trying to convert on. Draco mesh compression and KTX2/Basis Universal texture compression, both supported directly in the glTF ecosystem, typically bring that down by an order of magnitude without a visible quality loss (Khronos — glTF 2.0 Specification).

A model should load lazily, after the primary product image, and only initialize its viewer when it scrolls into view or the shopper explicitly requests the 3D view. Treating a 3D model as page-critical media the way a hero image is tends to be the single most common performance mistake teams make on their first rollout.

Failure modes teams hit on the first attempt

Missing usdz file: uploading only a .glb means iOS Safari has nothing to hand to AR Quick Look, so the AR badge silently never appears on iPhone traffic, which is often half the audience.
Wrong lighting bake: a model authored under studio lighting and dropped into a viewer with a generic environment map looks noticeably fake, which undermines the exact trust the feature was meant to build.
No accessible fallback: a 3D viewer with no alt text and no static image fallback fails for screen reader users and for any browser without WebGL, both of which are still real traffic.
Treating the model as decoration: a 3D asset with no measurement guide or scale reference gives a shopper a novelty, not the fit confidence the feature exists to provide.

Who actually owns this, operationally

3D asset creation is a production pipeline, not a one-time project. Most teams route it through whoever already owns product photography, since the input is usually the same physical sample, and the output needs the same quality bar and refresh cadence as a photo shoot.

Shopify's Partner Directory lists vendors who specialize in turning physical samples or CAD files into web-ready glTF and USDZ pairs, which is the workflow scaling teams tend to land on rather than building 3D capture in-house (Shopify.dev — Manage media for products and collections).

What is actually worth building today

3D and AR product viewing earns its cost on a narrow, specific category of product: anything where size, fit, or placement in a real space is the buyer's real hesitation. Furniture, large appliances, and anything sold by dimensions are the clearest cases.

The category that does not earn it is anything already well served by a photo and a size chart, apparel accessories and small goods among them. Building the pipeline for the whole catalog before proving it on a narrow slice is how these projects stall out.

A minimal rollout sequence

Pick 10-20 SKUs where physical scale is the actual purchase blocker, not the whole catalog on day one.
Get a glTF/glb model built or converted for each; most 3D asset vendors export both glb and usdz from one source file.
Add the model as product media through the platform's native 3D model type rather than a third-party embed, so it survives theme changes.
Verify the AR trigger separately on an actual iPhone (Quick Look) and an actual Android device (Scene Viewer), not just a desktop preview.
Measure product-detail-page conversion and return rate for the AR-enabled SKUs against a matched control group before expanding scope.

A 3D model that only renders correctly in a desktop preview and has never been opened on an actual phone is not shipped. AR viewing lives or dies on the device it is meant for.

None of this requires a metaverse platform, a wallet connection, or a new checkout flow. It requires two files per SKU and one web component, sitting inside the commerce stack that already exists.

FAQ

Is "the metaverse" completely dead for commerce?

Persistent branded virtual-world storefronts did not become a retail channel. Browser-based 3D and AR product viewing, which is a much narrower slice of the original pitch, did ship and stuck.

Do I need a Vision Pro, Quest, or other headset for this to work?

No. AR Quick Look and Scene Viewer both run through a phone's camera in the browser, with no headset and no app install required.

What file formats do I actually need?

A .glb file for Android and desktop viewers, and a .usdz file for Apple's AR Quick Look. Most 3D model vendors export both from the same source asset.

Does this require a new checkout or payment integration?

No. 3D and AR viewing is a product media type that sits on the existing product page; checkout and payments are untouched.

Which products actually benefit from this?

Products where physical size, fit, or placement is the real purchase hesitation: furniture, large appliances, anything typically returned over a size or scale mismatch.

Do we need to build a 3D capture studio in-house?

Most teams do not. Shopify's Partner Directory and comparable marketplaces list vendors who convert an existing physical sample or CAD file into a web-ready glb/usdz pair per SKU, which is cheaper than standing up in-house capture until volume justifies it.

References

From the Destm engineering archive. For current work on this topic, start at Solutions or the blog.