The problem AR solves in retail
Furniture is the hardest thing to sell online because the single question every buyer has — will it fit, and will it look right in my room? — cannot be answered by a photograph. That uncertainty is what drives both abandoned carts and returns, and returns on large items are brutally expensive once you count shipping both ways. An AR preview answers the question in the customer's own living room in about fifteen seconds.
Step one: finding the floor
Everything depends on the app knowing where the floor is. Native apps use ARKit on iOS and ARCore on Android, which fuse the camera feed with the phone's motion sensors to detect horizontal planes and keep tracking them as the user walks around. On the web, the modern path is WebXR hit-testing; where that isn't available, a well-built WebAR experience estimates the floor from the device's gyroscope and an assumed camera height, then raycasts onto that plane.
The detail most implementations get wrong is feedback. A phone pointed at a wall or held flat cannot find a floor, and an app that just spins a loader while the user waits feels broken. Good floor detection tells the user what to do — tilt down, move slowly, hold steady — and only confirms the plane once it has held a stable reading. You can see this working in our live room furniture demo: it asks you to point at the floor, shows a progress reading while it locks on, and then lets you tap anywhere to place a sofa, bed, or chair.
Step two: real-world scale
An AR preview is worthless if the object is not true to size. That means models must be authored in real-world units and checked against the actual product dimensions, not eyeballed. It also means resisting the temptation to let users freely resize furniture — scaling a sofa to make it "fit" defeats the entire purpose of the tool. Rotation and repositioning, yes; arbitrary scaling of a real product, no.
Step three: making it look like it belongs
Placement accuracy gets you halfway; believability gets you the rest. A contact shadow under the object anchors it to the floor — without one, furniture appears to hover. Lighting estimation matches the virtual object's brightness to the room. Occlusion, where supported, lets real objects pass in front of the virtual one. None of these are exotic features; they are the difference between "that's clever" and "that's what it would look like."
App or browser?
Native apps give you the best tracking and the deepest features, and they make sense when AR is part of a product customers already have installed. For everything else — a campaign, a catalogue page, a QR code on packaging or on the shop floor — WebAR wins on reach, because asking someone to install an app before they have bought anything is where most retail AR projects quietly die. A browser-based preview opens from a link and works on the phone already in the customer's hand.
What it costs
Two budgets matter, and they are separate. The application is largely fixed cost: a focused WebAR product viewer is a small project, while a native catalogue app with saved rooms, multi-item scenes and analytics is a multi-month build. The 3D assets are the variable: every product needs an accurate, optimised, real-scale model, and that cost scales with your catalogue, not with your app.
This is why we tell retail clients to start with their ten best sellers. Ten good models prove the conversion lift, establish the asset pipeline and standards, and cost a fraction of modelling everything. Our cost calculator gives a ballpark in a couple of minutes, or send us the brief and you'll have a fixed quote within 48 hours.
How we build them
We build AR product and furniture experiences with ARKit, ARCore, Unity AR Foundation and WebXR, including the 3D asset production in-house. Every engagement includes a fixed quote, an NDA and 100% source-code and asset ownership. See our AR development service, or try the live WebAR demos on your phone right now — no install.
Related: AR development · Unity app development · 3D production · Custom Quest environments
Quick answers ❓
How does an AR furniture app detect the floor?
Native apps use ARKit (iOS) and ARCore (Android), which combine the camera feed with motion sensors to find horizontal planes and track them as you move. Web-based AR uses WebXR hit-testing where available, or estimates a floor plane from the gyroscope and a known camera height. The app then raycasts from your finger onto that plane to place the object.
How much does it cost to build an AR furniture app?
A focused WebAR product viewer with a handful of models starts around $999–$3,000. A full native catalogue app with real-size scaling, saved rooms and analytics is a multi-month build quoted individually. Most of the variable cost is 3D asset production, not the app.
Do customers need to install an app?
No — WebAR runs in the phone browser. A customer taps a link or scans a QR code and the preview starts, which removes the biggest drop-off point in retail AR.
How many 3D models do we need to start?
Start with your ten best sellers. Proving conversion lift on a small set is faster and cheaper than modelling a whole catalogue, and it sets the pipeline standard for the rest.
Thinking about AR for your products? 🛋️
Tell us what you sell and we'll come back with an approach, a model-production plan and a fixed quote within 48 hours.