Newsroom

The lounge chair further down this page has ten fabrics and four frame colours, forty combinations in total. The client build it came from shipped those forty as forty separate 3D files of seven megabytes each, one fresh download per swatch click. It was rebuilt for this site as a single file that swaps materials on the spot.

That is the configurator most retailers get quoted, a 3D viewer with swatches wired to it, and it is the cheapest third of the job. The other two thirds are where the budget goes, and where the projects we are brought into have gone wrong.

The three jobs, in order of difficulty

Show the combination. Render what the shopper picked, accurately and fast, on a phone. For a fabric-and-finish product one model carries all the materials and a click swaps between them, with no second download. Every supplier can do that now; they differ on how true the fabric looks.

Price it. A running total that moves as the shopper chooses, with the surcharge on the swatch rather than in the basket. A delivery estimate sits next to it, because the lead time of a configured piece is the slowest option in it, not an average.

Hand a valid order to the back office. What the shopper built has to land in the order system as something a production planner can act on without phoning anyone. This job decides whether the configurator saves work or creates it, and it gets specified last, if at all.

Most projects arrive with job one done well, job two done partly, and job three handled by someone who reads the confirmation email and retypes it into the factory system. That person is the running cost of the configurator, and nobody put them in the business case.

Option rules break before the 3D does

The glTF extension that lets one model carry many material variants, the thing that turns the forty-combination chair into a single download, states that serving configurators is a non-goal: configuration has "a separate set of concerns" and is "more subject to complex application-specific business logic." The people who standardised the format looked at the rules and left them out deliberately.

The rules are therefore the retailer's problem. In furniture they recur in four forms.

Combinations that cannot be made. A chaise on both ends. A module count the frame will not carry. A fabric woven in the 140 cm width but not the 280 cm one, so it covers an armchair and not a four-seater. Constraints like these live in the heads of the people who build the product. Written into a PDF rather than into the tool, they will not stop it selling the impossible version.

Fabric families and grades. Upholstery is rarely priced per fabric. It is priced in grades, and a grade carries a surcharge, a supplier, a lead time and sometimes a minimum order. A configurator that models fabrics as a flat list of swatches has thrown away the structure the pricing depends on. We have seen that force a rebuild within a year, twice.

Lead times per option. Attach weeks to every option and show the maximum as the shopper chooses. Someone who learns at the confirmation email that their fabric takes eleven weeks either cancels or waits and resents the brand.

Options that no longer exist. A discontinued fabric still showing as a swatch is the most common fault in an otherwise finished configurator. The option list was exported from a spreadsheet once and never again.

None of this is 3D work. It is product data work with a 3D interface on top, and scoping it as a visualisation project is what puts these builds two or three times over budget.

The handover is where the platform fights you

Shopify raised its variant limit to 2,048 per product in October 2025, up from 100. The limit that bites is a different one: three options per product. The lounge chair, fabric plus frame, fits. A modular sofa with six layouts, forty fabrics, three leg finishes and an optional headrest is four option axes, and no variant allowance expresses that as one Shopify product.

So a real furniture configurator does not add to basket the ordinary way. It writes a line item with properties, or a draft order with a custom price, carrying the option codes for the system, a spec line a human can check, the configuration itself so support can reload what the customer saw, and the made-to-order and non-returnable flags.

Anyone evaluating suppliers should ask what the output looks like inside their own order system. If the answer is a string of codes, the budget needs to cover the person who translates it.

Build one family properly before the catalogue

Per-product pricing makes a big range look like a spreadsheet exercise, which is why the instinct is to quote the whole catalogue. The first family is where you find out that grade C is only available in three of the seven widths, and that nobody has a definitive list of which legs go with which frames. The second family costs a fraction of the first.

Pick one family, ideally the one generating the most option questions for customer service, and build it completely: rules, prices, lead times, the order payload. Then measure a month against the month before. Engagement, conversion on those SKUs, and the two numbers that pay for it: returns, and how many orders reached production without a phone call.

Online furniture comes back at around 22.7 percent, against 19.3 percent across all categories, and roughly 58 percent of furniture returns are size or space mismatches. A configurator carrying dimensions and web AR is going after the biggest reason things come back, not a marginal one.

Three of our clients went this way. Sweef started with the modular sofa, the product with the hardest combination question, using 3D resources they already had in house. MELIMELI led with the fabric browser beside the 3D sofa, fabric being the decision in that category. Contura built Build Your Stove from CAD files they already owned, and it ended up on the sales floor as well as the website.

Pricing is per product per month with modelling, hosting and yearly updates included, from 10 dollars per SKU on the Basic plan. Configurator work sits on the Value plan at 20 dollars per model per month. A large range gets an Enterprise quote.

To judge the modelling before committing to the rules work, send us the product that generates the most option questions. We will build it in 3D and AR for free and it is yours to keep. Send us one product.

Sources

• Khronos Group, KHR_materials_variants specification, including the non-goal statement on configurators.
• Shopify developer changelog, the product variant limit is now 2048 for all merchants, October 2025.
• Shopify Help Center, Adding variants, on the three options per product limit.
• Eightx, Furniture and home return rate benchmarks.
• Shopify, 3D ecommerce guide, for the configurator and made-to-order merchant results.