A person configures a car that does not exist yet. Wheel size, paint, stitching on the seats. Then they close the laptop. Research on SEAT’s configurator, published in the journal Forecasting in 2022, found that these people come back to the tool across a window of one to six months before they actually buy. The clicks arrive long before the money does. That gap is the real question. Making a product look good online is the easy half. The hard half is letting someone decide about a thing they cannot yet touch, and deciding is what the six-month window is full of.
IKEA Shipped Place in Seven Weeks
IKEA built its AR app in seven weeks and launched it on iOS 11 in the United States on September 19, 2017. The catalog at launch held more than 2,000 items: every sofa, armchair, footstool, and coffee table, plus part of the storage range. The app scaled each object to the dimensions of the room it was dropped into, with an accuracy IKEA put at 98 percent.
The speed was possible because IKEA had a head start. It had been a customer of Metaio, an AR startup Apple quietly bought in 2015, and that same team helped build ARKit. So when Apple opened the platform, IKEA already knew what it wanted to solve.
The problem was narrow and old. People buy a couch, get it home, and find it swallows the room. A tape measure tells you the couch is 210 centimeters wide. It does not tell you what 210 centimeters does to your living room. Place answered that exact question, and only that question. The furniture appeared at true scale on the floor in front of you, and the decision that used to require imagination now required a glance.
27 Percent, and 65 With AR
Rebecca Minkoff put 3D models on its product pages in the fall of 2019. The reason was mundane: the support team kept fielding the same questions about how bags looked and felt in the hand. Someone framed that as a hypothesis. Show the bag in three dimensions, and the questions should drop.
The numbers came back layered. Shoppers who interacted with a 3D model were 44 percent more likely to add the bag to their cart. Of those, 27 percent more went on to place an order. And shoppers who viewed a bag in AR, placed in their own space through the phone, were 65 percent more likely to buy.
Read those three figures in order and you see the mechanism. Each deeper level of interaction removes a specific doubt: what it looks like, then whether it is worth ordering, then whether it belongs in the room where it will live. Those are the doubts a salesperson answers one customer at a time. Here they get answered before anyone on the company side is involved. Bumbleride, which sells strollers, saw the same shape after adding 3D to its pages: conversion up 33 percent, time on the page up as much as 21 percent.
Safari Got a 3D Viewer in 2018
None of this needs a headset or an app. At WWDC in June 2018, Apple introduced AR Quick Look, a built-in viewer for 3D files in the USDZ format, which Pixar developed. From iOS 12 on, a USDZ file opens straight inside Safari, Mail, Messages, and Notes. The person taps once and the object is in the room.
A link is tagged as an AR object, the server delivers the right file, and the phone handles the rest. Before this, ARKit ran only inside apps, which meant every merchant needed its own app that a shopper had to download first. The browser erased that step. Shopify opened AR Quick Look to its merchants that autumn, and one of the first live storefronts running it was shop.magnolia.com. Compare the friction to the hardware path: Meta’s Quest Pro, shown that same autumn, cost $1,500.
This is where the choice of who builds the thing starts to matter, and it is worth being precise about what the work actually is. A polished render is table stakes. The harder job is accounting for the person inside the scene: where they will get lost, what they will fail to notice, which control they will never find, and what decision they walk away with. Judging that is closer to product work than to visual craft, which is why the studios that do it well tend to be the same ones recognized as the best website design company for this kind of build rather than for their portfolios alone. Google’s model-viewer, the open web component Shopify’s 3D pages run on, renders the same scene identically in Chrome, Safari, and Edge, so the render is the part that has already been solved. What has not is the person moving through it.
Horizon Worlds Missed 500,000
Immersion fails when it becomes the destination instead of the tool. Meta aimed for 500,000 monthly users in Horizon Worlds by the end of 2022, cut the target to 280,000, and still landed under 200,000, according to internal documents the Wall Street Journal reviewed that October. Only 9 percent of the worlds people built were ever visited by as many as 50 users. A great deal got constructed. Almost none of it was needed, which is the risk that grows as 3D gets cheaper to make.
Nike’s bet ran the same way. It bought the digital-sneaker studio RTFKT in December 2021, and by the end of January 2025 the studio had wound down, its site left as an archive. Worth separating these from the retail examples, though. Horizon Worlds and RTFKT were immersion sold as a product in its own right, not immersion layered onto a decision someone was already trying to make. Different category, different failure. What they show, read against the retail cases, is that a 3D space with no purchase decision attached to it drew fewer than 200,000 people even with Meta’s budget behind it.
Start With the Objections, Not the Technology
So here is the practical version, stated plainly. Immersive experience earns its cost when it is tied to a specific objection a buyer raises, and wastes it when it is added for effect. The way to tell them apart is not technical. Write down every question your sales team answers by hand, every reason a deal stalls, every thing a customer cannot picture and therefore will not commit to. That list is the brief. If a 3D scene or an AR view removes an item from it, build that. If it removes nothing, you are building a demo.
The real-estate data shows how far a behavior can move when the objection is the right one. Before the pandemic, listings with an open house outnumbered listings with a 3D tour by 8.2 to 1. By April 2021 that ratio had fallen to 1.6 to 1. The tour was not decoration. For a house someone could not walk through, it was the only way to see the thing at all, and buyers reorganized their behavior around it within a year. In the same stretch, from March 2020 to February 2021, Zillow reported that listings carrying one of its 3D tours were favorited 75 percent more often than listings without.
Back at the configurator, the loop closes in a way that is easy to miss. Those clicks, spread across the one-to-six-month window, turned out to predict the weekly color mix of actual sales more accurately than past sales alone, by up to 25 points in the study’s metric. Color, because it can be changed almost to the last step of the production line. The person building a car that does not exist yet was not only deciding what to buy. They were telling the factory which paint to load.
