Furniture retail has carried the same two questions since the first showroom opened: will it fit, and will it match. Every product photo dodges both. The rug looks beautiful on the photographer's parquet, at the photographer's angle, in the photographer's light, and none of that is your living room. So customers guess, and a meaningful share of them guess wrong, and wrong guesses come back in a truck.
A 3D room planner attacks both questions directly: an interactive room in the browser where customers place real products, at real sizes, on floors they can change to resemble their own. I built one for Bizsan, a Hungarian rug retailer, and this article is a walk through what these tools actually are, what made that one work, and what a business should know before commissioning something similar.
A planner is not a configurator
The distinction matters for scoping. A product configurator answers questions about one product: which finish, which size, which options. A room planner answers questions about a space: what belongs in it, how pieces work together, whether that 170 by 240 rug actually fits between the sofa and the wall. The configurator is a product page with depth. The planner is closer to a very small game, with a camera to steer, objects to arrange, and a scene that has to stay coherent while the customer rearranges it.
That makes planners more ambitious builds. There is a furniture library to manage, collision handling so objects do not overlap or float, floor and wall materials to switch, and a camera that has to feel effortless to someone who has never controlled a 3D scene in their life. In the Bizsan planner that last problem got a deliberately humble solution: an on-screen joystick, borrowed straight from mobile games, because it is the one navigation control most people have already used.
The decision that made it a sales tool
The single most important architectural decision in that project had nothing to do with rendering. The planner does not use a hand-picked demo set of products. It talks to the shop's live catalog through an API: real products, real prices, real stock status, real dimensions, straight from the store database. Choose a rug in the planner and you are looking at an actual SKU, with a link back to its product page where you can buy it.
This is the line between a marketing toy and a sales channel. A demo scene with ten beautiful rugs impresses visitors once. A planner wired into the catalog sells whatever the store sells this week, updates when inventory changes, and never needs a developer to add a product. When the shop adds a new collection, the planner has it, because the planner and the shop read from the same source of truth.
Turning product photos into 3D textures, automatically
The obvious blocker for any retailer with a large catalog: where do the 3D assets come from? Nobody is going to hand-model four hundred rugs, and nobody is going to reshoot four hundred product photos on a calibrated rig either. The budget dies right there in most conversations.
For rugs, it turned out the product photos the shop already had were enough, with processing. The planner runs each catalog photo through an image pipeline in the browser: it samples the photo's background color, measures how far every pixel is from that background in color and brightness, strips the background away, and cleans the edges. What is left is the rug itself, which becomes a texture on a correctly sized plane in the scene. Several hundred products became placeable 3D objects with zero reshoots and zero manual asset work per product.
The honest caveat: this works because rugs are flat. A sofa needs a real model, and that is the main cost driver separating a rug planner from a full furniture planner. But the principle transfers to any business: before assuming you need custom 3D assets for everything, look hard at what your existing product data can be converted into. The answer decides most of the budget, which is exactly the pattern I described in what 3D websites cost.
Scale honesty
The quietest feature is the one doing the most commercial work. Every rug in the planner is placed at its true dimensions, pulled from the same product attributes the shop displays. When a customer drops a 200 by 300 into the room and it dwarfs the sofa, that is not a rendering aesthetic, that is the purchase decision happening early, in the browser, instead of late, in their living room with a return label.
Returns are where the business case usually closes. Furniture and rug returns are expensive in shipping, handling, and re-stocking, and size or fit disappointment is a leading reason for them. A tool that moves the fit question before the purchase does not need to be dramatic to pay for itself. It needs to prevent a modest number of wrong-size orders per month.
What else lives inside one of these
- A furniture library, so the customer can stage the product among believable neighbors instead of judging it in a void.
- Floor and wall material options, because a rug reads completely differently on light oak versus dark tile, and the customer knows which one they have at home.
- Physics for placement, so objects rest and collide naturally instead of intersecting like ghosts. It sounds cosmetic, and it is actually trust: the first time a chair sinks through a rug, the customer stops believing the room.
- A guided tour for first-time visitors, three or four hints, because the audience for a furniture planner is not gamers and the tool has one chance to feel approachable.
- Careful camera limits: a camera that can end up under the floor or inside a wall will, for some customer, within the first hour of launch.
What the business actually gets
Beyond the returns argument, the measurable effects cluster in three places. Session time rises, because arranging a room is inherently engaging in a way that scrolling a category grid is not, and the research on interactive commerce backs the pattern: interactivity produces presence, and presence produces purchase confidence. Differentiation is real in categories where every competitor has the same grid of photos on white. And there is a data channel nobody expects: the planner sees which products get placed, in which combinations, on which floors, which is merchandising intelligence a photo grid can never produce.
One more effect deserves mention because clients keep reporting it: the planner works as an in-store tool as much as an online one. A tablet on the showroom floor running the same planner lets a salesperson stage any product from the full catalog in seconds, including pieces not physically present. The website feature quietly becomes sales infrastructure.
What it takes to build
The Bizsan planner runs on React Three Fiber with a physics layer for placement and a post-processing pass for the look, wrapped in interface panels for the catalog, furniture, and room controls. The details matter less than the shape of the work: roughly a third of the effort went into the scene itself, a third into catalog integration and the photo-to-texture pipeline, and a third into interface and approachability, the panels, the joystick, the tour, the hundred small decisions that decide whether a non-technical customer stays past the first ten seconds.
From the business side, the requirements are less technical than organizational. The planner is only as good as the product data feeding it, so dimensions have to exist and be correct for every product, photos have to be reasonably consistent, and someone at the company has to own answering questions quickly. Every planner project I have seen stall has stalled on data, not on rendering.
Budget-wise these land in the upper configurator range or above, depending on catalog size and how much of the asset pipeline can be automated. Phasing works well: a single-room planner with one product category proves the concept and the numbers, and the furniture library, additional rooms, and AR ambitions can arrive as later phases on the same foundation.
The short version
- A room planner answers the two questions photos cannot: will it fit, and will it match.
- Wire it into the live catalog or accept that you built a demo. Real SKUs, real prices, real links to buy.
- Audit your existing product data before budgeting custom 3D assets. Sometimes the photos you have are the assets.
- True-to-scale placement is a returns-reduction feature dressed as a visual.
- Budget a third of the project for approachability. The scene is for the customer who has never touched 3D.
The full case study of the Bizsan build, with screens of the catalog integration and room controls, is in my projects: Bizsan 3D Room Planner. And if you are weighing whether your catalog and your customers would justify one, that is exactly the kind of conversation worth having before any code exists.