3D for e-commerce6 min read
A manager's guide to web AR: what it needs, what it costs, how long it takes
Web AR explained for e-commerce managers: what it actually is, what your team has to provide, realistic timelines and cost, and how to judge a vendor.
By CharpstAR · 25 Sept 2023 · Updated 22 Sept 2026
Part of Augmented reality for furniture retail: the complete guide
If you manage an e-commerce site, web AR is probably on a list somewhere with a vague note attached. This is the explanation a manager actually needs: what it is, what it asks of your team, how long it takes, what it costs, and how to tell a serious vendor from a slide deck.
What web AR actually is
Web AR means a shopper taps a button on your product page and the product appears in their camera view at real size, with no app installed. It works because both mobile operating systems have this built in.
On Android, the button hands a glTF 2.0 or GLB file to Google's Scene Viewer, which needs an ARCore supported device running Android 7.0 or later with a current Google Play Services for AR. On iOS and iPadOS, it hands a USDZ file to AR Quick Look, which has shipped since iOS 12 and works in Safari.
That is the whole mechanism. Your site does not run the AR session; the phone does. Your site provides a file and a link.
The same model, in a browser viewer, also gives you the thing most shoppers actually use: a product they can rotate on any device, including desktop.
What it needs from your team
This is the part vendors under-describe. Web AR is not a plugin you switch on.
Product data. Accurate dimensions for every product you want modelled, in a consistent unit. If your dimension data is wrong, the AR experience is wrong, and wrong at real size is worse than no AR at all.
Reference material. Photographs from several angles including the back and underside, material and finish information, and CAD if you have it. Products with CAD are faster and cheaper.
A named person who can approve. Someone who can look at a model and say "the frame is too thick" or "that oak is too orange". Review is a real workload, a few minutes per product across one or two rounds. If nobody owns it, the project stalls at 80 percent finished.
Front-end access. Someone who can add a viewer component to the product page template. On Shopify or similar this is usually small. On a bespoke front end, budget a developer week or two for the template, poster handling, fallback behaviour and analytics events.
A realistic timeline
For a first phase covering 50 to 100 products, this is what we typically see.
| Stage | Duration | Who does it |
|---|---|---|
| Scoping, product selection, data check | 1 to 2 weeks | You, with the vendor |
| Reference gathering for the first batch | 1 week | Your team |
| First models delivered for review | 1 to 2 weeks after references | Vendor |
| Review and revision rounds | 1 to 2 weeks | You |
| Front-end integration on the product page | 1 to 3 weeks, in parallel | Your developers |
| Remaining models, in batches | 3 to 8 weeks | Vendor |
| Live, measuring | ongoing | You |
So six to twelve weeks from first call to a meaningful number of products live, with the front-end work overlapping the modelling. The two things that stretch it are missing reference material and an unstaffed review step. Anyone promising a thousand products live in a month is either automating the modelling in a way that will show, or has not read your catalogue.
What it costs
There are two cost models and you should know which you are being sold.
Per model, one off. You pay a fee per product for a model you own. Rates vary enormously with complexity, from a rigid box to a configurable upholstered sofa. You then own the files and the hosting problem.
Per product per month. A subscription covering modelling, hosting, format conversion, the viewer and updates. Our own pricing works this way: Basic starts at $10 per product per month on a 12-month term with a minimum of 100 variants, Value starts at $20 per model per month on 24 months, and Enterprise is quoted.
Subscription usually wins when your catalogue evolves, because models need maintaining and re-exporting as formats change. One-off wins with a stable range and in-house 3D capability.
Either way, add the internal cost: review time, developer time and analytics. Projects that budget only the vendor invoice are the ones that run late.
What you should expect it to do
The published evidence is decent but it needs reading carefully.
Shopify reported in 2020 that interactions with products carrying 3D or AR content converted 94 percent more often than the same kind of products without. Shopify has also published Rebecca Minkoff results: shoppers were 44 percent more likely to add an item to cart after viewing it in 3D, 27 percent of those were more likely to place an order, and AR-enabled product pages saw a 65 percent purchase lift. Oakywood reported average order value up 458 percent in Germany, 232 percent in Switzerland and 84 percent in the UK after adding Shopify AR and 3D customisers.
Read all of those as strong signals, not forecasts. They come from merchants who chose which products to model, and who tend to improve the rest of the page at the same time. Plan against the number you generate yourself, comparing conversion for sessions that interacted with the model against sessions that did not.
On returns, be conservative. The National Retail Federation and Appriss Retail put United States returns at 743 billion dollars in 2023, 14.5 percent of sales and 17.6 percent online. AR reduces one slice of that: returns caused by the product being the wrong size for the space. It does nothing for colour disappointment, quality complaints or late delivery.
How to judge a vendor
Can I see a model of a product like mine, live on a phone, right now? Not a video, not a render. A link. If they cannot produce one in the category you sell, you are the pilot.
What file sizes do you deliver? Google recommends under 10 MB with a hard limit of 15 MB, textures at 2048 by 2048 or below, and 30,000 to 50,000 triangles as the ideal range. Shopify puts the ideal GLB at around 4 MB. A vendor who does not know these numbers has not shipped to a real product page.
Do I own the models? Ask what happens to your files if you leave, and get it in the contract.
Who does the format conversion and hosting as formats change? GLB and USDZ are both moving targets. Somebody has to maintain this.
How many revision rounds are included, and what counts as a revision? Budgets break here.
What do you need from me, in a list? A vendor who cannot immediately name the reference material they need per product has not thought about your side of the work.
Add a seventh: ask what their models cannot do. A vendor with no honest limitation to offer is selling, not advising.
Where to start
Pick ten products that are large, expensive and hard to judge from photographs. Put them on live product pages with proper analytics, compare against the rest of the catalogue for a month or two, then decide whether to extend.
That is what our own clients did. Sweef went from a pilot to a full modular sofa configurator. Contura started with the stoves where clearance questions cost the most support time. MELIMELI reuses the same models across shop and campaign material rather than reshooting.
If you want to run that pilot without a purchase order, we will build a 3D model of one of your products free. Send us one product and the model is yours to keep and put live. What we build and what it costs after that is on solutions and prices.
Sources
- Google, Scene Viewer documentation.
- Apple, AR Quick Look.
- Shopify, 3D e-commerce.
- Shopify, merchant data on 3D and AR content, September 2020.
- Shopify, AR shopping.
- National Retail Federation and Appriss Retail, 2023 Consumer Returns in the Retail Industry.
