The store still sells most sofas, and the page sells the rest
35%
of people who bought a sofa or loveseat in the past year bought it online. The other 65% in a store.
Provoke Insights, Furniture Trends, n = 1,501 US adults aged 21 to 65, fieldwork September to October 2024Most sofas are still bought in a store. In Provoke Insights' survey of 1,501 Americans aged 21 to 65 in autumn 2024, 65% of the people who had bought a sofa or loveseat in the past year bought it in a store. For sectionals it was 69%.
The rest is not small. 35% bought their sofa online. Across all furniture, 784 of the 1,501 people in the same survey had bought furniture online in the past year.
The store wins, and I think for a good reason. A showroom answers three questions before anyone pays: will it fit, what does the material really look like, and how does it work. A product page that leaves those open gets them answered after delivery.
Why it comes back: five reasons, in the order a buyer meets them
44%
of people who returned something said it didn't meet expectations for quality, colour or size
YouGov, n = 1,311 US adults, 11 and 12 September 2025Online, about a fifth of sales comes back. The National Retail Federation and Happy Returns estimated that 19.3% of online sales would be returned in 2025. That covers all of retail. We haven't found a reliable figure for furniture alone, so we don't quote one.
The reasons are better documented than the rate. In YouGov's survey of 1,311 American adults in September 2025, 44% of the people who had returned something said it didn't meet their expectations for quality, colour or size. 39% named poor fit. 34% named an item that arrived damaged or defective. People could give more than one reason, so the numbers don't add up to 100.
For a sofa, a table or a bed, those reasons sort into five groups:
- 01It doesn't fit the room, the doorway or the stairwell. The buyer read the dimensions and still got the scale wrong.
- 02The material looks different in the buyer's own light than it did on the page. Fabric, leather or wood.
- 03It doesn't work the way the buyer pictured. The chaise is on the wrong side, the module doesn't detach, the drawer opens the other way.
- 04It arrived damaged, or so late that the buyer had already bought something else.
- 05The buyer changed their mind. No product page prevents this, and it is not worth fighting.
The three a product page can fix are the showroom's three questions
Will it fit. Baymard Institute tested furniture and home decor sites, IKEA, Wayfair and Pottery Barn among them. Their finding: "Most users uncertain about the dimensions will abandon the product, while others will misjudge the fit, increasing the likelihood of a return." The fix they recommend is to link each measurement visually to the part it describes. A real-time model does that by nature: the measurement stays on the edge it measures while the buyer turns the sofa. On a phone, AR puts the piece on the buyer's own floor at true size, so the doorway question gets answered before the order.
What the material really looks like. A photograph shows fabric in one light, chosen by the photographer. A model shows it from every angle the buyer picks, and the weave or the grain reacts as the buyer moves. This does not replace your photography. Photography sets the mood. The model answers what comes next.
How it works. Whatever moves on the piece can be moved on the page. A buyer who has moved the chaise to the left side themselves, and watched the module come apart, doesn't order the mirror configuration by mistake.
The page answers the showroom's three questions for the buyer who isn't coming in.
The two it cannot fix, and why it's worth saying so
Damage in transit is a packaging and carrier problem. In the same YouGov survey, 34% of the people who returned something named an item that arrived damaged or defective. A page has nothing to do with that. If damage is a large share of your returns, fix the packaging before you spend on the page.
Change of mind is the cost of selling something people live with for ten years. The page can shorten the hesitation before the order. It can't remove the right to change one's mind, and it shouldn't try.
Where 3D and AR make returns worse
9%
narrower and shallower than the specification, on a two-seater we built. Height was exact.
Our own build, measured against the artist brief
Important to note, because vendor pages leave this part out: a model at the wrong scale does more damage than no model at all, because AR shows it to the buyer as fact. The buyer puts the sofa on their floor, sees it fit, orders it, and the real one doesn't.
We caught this in our own work, and I'm glad we did. On a two-seater we built this year, the specification we gave the artist said 138 by 68 centimetres. The model that came back measured 125.5 by 61.8. The height matched to the millimetre, which is why nobody noticed by eye. In AR that sofa would have stood 12.5 centimetres narrower than the one in the showroom. We found it while putting the specification's dimension lines around that model on our own site, and the numbers didn't match what we saw.
What we do now: before a model goes live, every axis is measured against the specification, and the number displayed on the page is read from the specification, never from the model, because a beautiful wrong model raises returns.
The second way it goes wrong is a texture that's prettier than the fabric. A weave rendered cleaner than it is in the room, or leather without its grain, sets an expectation the delivery can't meet. The model has to show the material as it is, including the parts that are not pretty.
Every axis of the model is checked against the specification before it goes live. The dimension on the page is read from the specification, never from the model.
What to measure before you build anything
If I had to pick one number, it would be this one. Ask your returns team what share of returns are tagged as 'not what I expected', in any wording. Not the overall return rate. That share is the part a page can move, and it's the baseline you compare against three months after the model is live.
If that share is small and damage is large, spend on packaging. If it's large, the page is the right place to spend, in the showroom's order: will it fit, then what the material really looks like, then how it works. That is also the order of cost.
The share of returns tagged 'not what I expected' is the only number worth measuring before a model, and the one to compare three months after.
What to take away
A showroom answers three questions before the sale: will it fit, what the material really looks like, how it works. Online, the product page has to answer the same three, or the return answers them. Damage and change of mind it cannot fix. One check before anything goes live: every axis of the model against the specification. We caught a 9% width error on our own build, and AR would have shown it as fact.