ARKit and the AR moment that almost arrived.

Every brand meeting since October has had AR on the agenda. Apple's ARKit landed with iOS 11 in September; Google's ARCore went stable in February. Between them they've put a usable, marker-less, six-degrees-of-freedom AR runtime on hundreds of millions of phones in the wild, and the marketing teams have noticed. Every brand we work with has now asked the same question in some form: 'what should we be doing with AR?'.

We've spent the last six months saying 'maybe' to a lot of pitches, 'yes' to a few, and 'absolutely not' to several. Here's the picture as we see it.

The state of the platform

What's actually working

• 01Furniture and product preview — IKEA Place, Wayfair, Made.com. Place the thing in the room, decide. The use case is clear, the value is obvious, the friction is low.
• 02Try-on for cosmetics and eyewear. Useful, repeat-use, low gimmick. Sephora and Warby Parker show the pattern.
• 03Wayfinding overlays for events and venues — where it's well-instrumented. We've prototyped one for a museum brief that genuinely earned the budget.
• 04Industrial maintenance overlays — show the engineer which valve to turn, on the actual valve. Niche but real.
• 05Children's books and educational toys — the camera reveals an interactive layer over the printed page. Limited commercial reach but high delight.

What isn't

• 01AR 'experiences' that launch from a print ad. The friction (download app, point camera, hold steady, wait for tracking) kills it before the payoff arrives.
• 02AR games that require a flat, well-lit, table-sized surface. Real homes don't have those.
• 03Any AR feature where the user has to install an app just for the AR feature. Almost nobody does.
• 04Anything that requires the user to hold their phone in front of their face for more than 60 seconds. Arms get tired; users get bored.
• 05Multi-user AR experiences in 2018 — the shared session APIs are too immature for production use.

"

AR is at the same point in the hype cycle that VR was in 2014. Real platforms, real demos, no real consumer behaviour yet. The brands that win the AR conversation in 2020 are the ones whose 2018 work shipped quietly and earned repeat use.

The 'experience' problem, explained

Most AR briefs we receive describe an 'experience' — a one-time interactive moment, often tied to a campaign or launch. The brand expects users to download a dedicated app, point their camera at something, and engage for a few minutes. The metric is press coverage; the success criteria are vague.

Here's what actually happens. A small percentage of campaign-aware users download the app. A subset of them actually open it within 24 hours. A subset of those make it through the AR tracking onboarding. A subset of those see the experience. The funnel typically loses 95% of the engaged audience between 'aware of the campaign' and 'completed the AR moment'. The remaining 5% are not worth the production cost.

Web-based AR (AR Quick Look on iOS 12, hopefully later this year) will close the install gap and is the next thing worth watching. Until it ships, single-use AR campaign apps should be off the table.

Our advice to clients in 2018

• 01Build AR when it removes a real friction — 'will this sofa fit?', 'how do I look in these glasses?', 'how do I install this thing?'
• 02Don't build AR because the brief said 'innovation'. The platform is real, but the use cases are narrower than the hype suggests.
• 03If you must build a campaign-driven AR moment, wait for web-based AR. The install funnel will kill anything else.
• 04Test with real users. AR demos look excellent in the studio. They behave very differently on a customer's phone, in a customer's lounge, holding the camera at the wrong height.
• 05Budget for content, not just code. The 3D models for a useful AR product preview cost as much as the app that displays them — sometimes more.

What we think happens next

• 012018–19: Web-based AR (AR Quick Look on iOS, Scene Viewer on Android) becomes the default delivery for product preview use cases. App-store-AR becomes niche.
• 022019–20: AR glasses move from 'always next year' to 'announced but underwhelming'. Apple is rumoured to be working on hardware; nothing ships in this window.
• 032020–22: Genuine usefulness for AR settles around three use cases — product preview, navigation, and industrial overlays. Everything else fades.
• 04Beyond 2022: AR becomes glasses-shaped or doesn't really happen at consumer scale. The phone-camera version is a transitional form factor.

ARKit and ARCore are real platforms; they were not vapour. The hype around them, today, is unbalanced relative to the demonstrated user behaviour. The right way to engage with AR in 2018 is targeted, useful, low-friction, and resistant to the 'wow factor' brief. The brands who'll dominate AR in five years are mostly the ones quietly shipping useful product-preview features today, not the ones running launch-week press hits.