Guides6 min read
How to run your first 3D product page pilot in 30 days
A practical plan for a first 3D and AR pilot: which ten products to pick, what to put in the brief, and the four numbers to measure before you scale.
By CharpstAR · 4 Oct 2023 · Updated 22 Sept 2026
Part of 3D models for e-commerce: the complete guide
Most 3D projects die at the same point. Someone sees a demo, gets excited, asks for a quote for the whole catalogue, and the number comes back large enough that the idea moves to next year's budget. A pilot avoids that. Ten products, thirty days, four numbers at the end, and a decision made with your own data rather than with a vendor's case study.
This is the plan we give clients who have never had a 3D model on their site. It assumes you have a normal e-commerce team, no 3D skills in house, and no appetite for a six-month project.
Week one: pick the ten products
The selection is the part people get wrong. The instinct is to pick the ten best sellers. That is usually the worst choice, because best sellers already convert well and there is little room left to show a difference.
Pick instead on these criteria:
- Size matters to the buyer. Sofas, stoves, desks, wardrobes, large lighting. If the shopper's real question is "does this fit", a scaled 3D model answers it and a photograph cannot.
- The product has variants. Fabrics, finishes, modules, sizes. One model with swappable materials replaces a photo shoot per combination, so the cost per visible variant collapses.
- Photography is currently weak. Products shot on a white background from three angles, where the back, the underside and the joinery are invisible.
- Traffic is enough to measure. A page with 200 sessions a month will tell you nothing in thirty days. Aim for products that together see a few thousand sessions a month.
- Returns are painful. If a product has a return rate well above your average and the reason codes say size or expectation, it belongs in the pilot.
Ten is the right number because it is enough to see a pattern and small enough to finish. Two or three products gives you an anecdote.
Week one, also: write a brief that a modeller can actually use
A good brief prevents the two-week revision loop that ruins most pilots. What a 3D team needs from you:
- Dimensions from the spec sheet, not from the marketing copy. Height, width, depth, seat height, plinth height.
- CAD files if they exist. Most furniture and appliance brands have them from their manufacturer and do not think to offer them. They cut modelling time and remove guesswork on proportions. Contura's stove models were built from CAD for exactly this reason.
- Photographs from every side, including the back and the underside, plus a close-up of any joint, seam, hinge or stitching.
- Material references. Fabric swatches photographed flat under even light, the wood finish, the metal finish, and the paint code if you have one.
- One reference image of how it should look when finished, so there is no argument about lighting style later.
Say clearly which variants must be switchable in the model and which are a separate model. That single decision drives most of the cost.
Week two and three: production and placement
Production of ten models typically runs in parallel and comes back as GLB files for the web and USDZ for iPhone. You do not need to care much about the formats beyond knowing both exist. Android devices use Scene Viewer, which reads glTF 2.0 and GLB, and Apple devices use AR Quick Look, which reads USDZ. Shopify's own guidance is that GLB and USDZ are the supported formats, that files over 15 megabytes get optimised automatically, and that around four megabytes is a sensible target.
Placement matters more than most people expect. Put the viewer where the main image is, or immediately beneath it. Not in a tab labelled "3D" at the bottom of the page, which is where these things go to be ignored. On mobile, the viewer should be visible without scrolling past the price.
Keep the AR button present but secondary. Rotation is the majority behaviour, AR placement is the minority one, and a page designed around the minority behaviour underperforms.
Set up measurement before the models go live, not after. This is the second most common way pilots fail: the models ship, everyone likes them, and nobody can say what happened.
Week four: the four numbers
Track exactly four things and resist the urge to add more.
| Number | How to get it | What good looks like |
|---|---|---|
| Viewer engagement rate | Fire an event on first rotation, divide by sessions that scrolled the viewer into view | A solid majority of people who see it will touch it |
| AR open rate | Event on AR button tap, divide by mobile sessions | A minority behaviour; treat it as a bonus, not the point |
| Conversion, page versus control | Compare the ten pilot pages against ten matched pages you did not change | Any consistent direction across ten products |
| Return rate on pilot products | Your existing returns report, filtered to the ten SKUs | Needs longer than 30 days; start the clock now |
The control group is the part people skip. Without ten comparable products left untouched, a seasonal swing or a promotion will make your pilot look brilliant or terrible for reasons that have nothing to do with 3D.
Be honest about what thirty days can and cannot tell you. Engagement and AR open rate are readable inside a month. Conversion needs enough orders to be meaningful. Returns need a full return window plus processing, so realistically a quarter.
What the published numbers suggest, and how to read them
Shopify reported in 2020 that interactions with products carrying 3D or AR content converted 94 percent more often than products without. That is a comparison across merchant data, not a controlled test, and merchants tend to model their better products, so read it as a signal rather than a forecast. The Rebecca Minkoff figures are more concrete: shoppers 44 percent more likely to add to cart after viewing a 3D model, 27 percent more likely to order, and a 65 percent purchase lift on AR-enabled pages.
Your pilot exists to produce your own version of those numbers. If it produces something smaller, that is still useful information, and it is worth far more than a slide.
What it costs to keep going
If the pilot works, the question becomes how to scale without a capital project. Our own pricing is built for that: Basic starts at 10 dollars per product per month on a 12-month term with a minimum of 100 variants, and Value starts at 20 dollars per model per month on a 24-month term. Enterprise is quoted. The point of a subscription rather than a per-model invoice is that the pilot cost stays small enough to approve without a business case.
Three clients took this route. Sweef started with internal 3D resources and ended with a customer-facing modular sofa configurator and web AR. Contura built a stove configurator from CAD files. MELIMELI brought the in-store sofa experience into their online journey.
If you want to start smaller than ten products, start with one. We will build a 3D model of a single product for free so you can put it live and watch what happens. Send us one product and the model is yours to keep. What we build after that is described on solutions.
Sources
- Shopify, merchant data on 3D and AR content, September 2020.
- Shopify, AR shopping and 3D e-commerce.
- Google, Scene Viewer documentation.
- Apple, AR Quick Look.
