Ask a hundred people to measure their living room and you will get a hundred different floor plans. Not because people are careless, because rooms are genuinely awkward. Skirting boards eat centimetres. Radiators stick out. Bay windows refuse to be a rectangle. And a tape measure held by one person across four metres sags in the middle.
That gap between the room someone thinks they have and the room they actually have is where furniture returns come from.
The number nobody checks twice
Retail teams have known this for a long time. When a customer buys a three-seater, the decision usually rests on a single measurement taken once, written on a phone note, and never verified. It feels precise. It isn't.
A five-centimetre error sounds harmless until it is the difference between a sofa that fits the alcove and a sofa going back on a van.
The cost is not just the refund. It is the second delivery slot, the restocking, the item that now cannot be sold as new, and a customer who will hesitate before ordering anything large again.
What goes wrong, specifically
- Diagonal drift. Tape measures are pulled at an angle far more often than people realise, and the error always runs in the same direction, too long.
- Obstruction blindness. Sockets, pipe boxing and door swings are rarely recorded, then discovered on delivery day.
- Unit slips. Mixing feet and centimetres inside one plan is more common than any of us would like.
- No second opinion. There is nothing checking the numbers before they turn into an order.
Why phones changed the maths
Modern phone sensors, depth cameras, LiDAR on higher-end devices, and increasingly good monocular depth models, can reconstruct a room's geometry from a slow walk-through. The measurement stops being a single human act and becomes thousands of samples averaged into a surface.
That matters for one reason: errors stop compounding. A tape measure has one chance to be right. A scan has ten thousand, and the wrong ones cancel out.
Where accuracy actually lands
In our testing across flats, period houses and open-plan extensions, room dimensions come back inside roughly a centimetre against a laser reference. That is more than good enough for furniture, flooring quantities, and paint coverage, the three decisions that drive the most expensive mistakes.
The interesting part is not the accuracy figure. It is that the measurement becomes reusable. Scan once, and every subsequent decision, the rug, the shelving, the flooring order, inherits a room that is already correct.
What we would tell a retail team
1. Treat the measurement as part of the product experience, not a chore you hand to the customer.
2. Capture obstructions, not just walls. The alcove depth matters more than the room length.
3. Store the plan. A customer who scanned once will come back to buy against it.
The room is the constraint every other decision hangs off. It is worth getting right first.
What people do instead, today
In the meantime, the workaround most people actually use is a tape measure, a rough sketch, and a lot of "that should fit, probably." It mostly works. It also produces the exact failure mode that makes measurement worth solving properly: the wardrobe that's 4cm too deep for the alcove, discovered after delivery, not before.
Professional measurement, a surveyor, a fit-out company's own team, solves this reliably, at a cost and a lead time that doesn't make sense for "will this sofa fit." That gap, between a tape measure and a professional survey, is where most furniture-buying decisions actually happen, uncertainly.
Why a photo alone doesn't solve it
A single photo tells you a lot about a room's style, enough for [a full redesign](/blog/how-does-ai-interior-design-work), but comparatively little about its precise dimensions. Perspective distorts distance in ways that are easy for a person to misjudge and genuinely hard for software to correct for without more information than one image provides: a second angle, a known reference object, or ideally a short walkthrough video rather than a single frame.
That's the real reason measurement and redesign are separate features here rather than one. Redesign needs to know where your walls are, relatively, enough to keep them fixed while changing what's on them. Measurement needs to know how far apart they are, in centimetres, precisely enough to bet a sofa purchase on. Those are different problems with different error tolerances, and conflating them is how you end up with a tool that's mediocre at both instead of good at one.
