Augmented reality layers digital content onto the physical world, and most smartphone users have a device capable of AR in their pocket. Every recent iPhone and most mid-range Android phones include a depth-aware camera, a tracking stack (ARKit on iOS, ARCore on Android) and enough GPU to render AR overlays at 60 fps. That changes the calculus for what is worth building: the question is no longer "can our users run this?" but "is there a moment in their day where an overlay genuinely helps?"
Beyond the novelty filter
The early wave of AR was face filters and furniture previews. They were primarily novelty experiences with limited repeat usage. The current wave is practical: navigation overlays that show the turn on the pavement in front of you, live product information when you point a camera at a shelf, guided repair that highlights the exact screw to remove, and try-before-you-buy that measurably reduces returns.
The difference between the two waves is not technology. It is intent. A filter exists to be looked at; a good AR tool exists to be looked through. The camera view is the interface, and the digital layer answers a question the user already had — where do I go, will this fit, which part is broken, what does this cost. When the answer arrives faster than it would have by any other route, people come back.
The web is an AR platform now
WebAR means you can deliver an experience through a link — no app install, no friction. Browser support for WebXR has matured on Android, and on iOS the practical route is Quick Look for 3D models plus camera-based libraries for tracking. Either way, a customer can tap a QR code on packaging, or a button on a product page, and see the thing in their room within seconds.
For marketing and commerce this removes the single biggest reason AR projects used to fail: the install. Asking someone to download a 200 MB app to preview a lamp is a request most people decline, and the conversion data reflects it. A link they can open in the browser they already have gets used, and it gets shared — which turns a one-off campaign into something that spreads on its own.
The trade-off is capability. WebAR cannot do everything a native app can. Tracking may be less stable, lighting estimation less precise, and performance can vary on older devices. The right pattern for most businesses is to start on the web, measure whether anyone uses it, and only invest in a native experience once there is evidence that the moment is real.
Where AR is generating real value
A few core use cases generate the majority of ROI we observe in production:
- Product visualisation in a real space. Furniture, lighting, appliances, flooring and paint. The value is not the wow — it is that a customer who has seen the sofa in their living room is far less likely to send it back.
- Step-by-step guided instructions. Assembly, installation, maintenance. An overlay that highlights the next component beats a paper diagram and a support call, and it removes the ambiguity that causes most errors.
- Contextual information overlays. Field technicians pointing a phone at connected equipment to see its service history, warehouse staff seeing pick locations, engineers seeing what is behind a wall before drilling.
- Interactive packaging and print. A label or poster that becomes a video, a 3D model, or a configurator when scanned. Cheap to produce and easy to measure, because every scan is an event you can count.
- Remote assistance. An expert in one place annotating what a person in another place is looking at. This is where AR quietly pays for itself in industrial and healthcare settings, far from the consumer spotlight.
Headsets change the ceiling, not the floor
Apple Vision Pro, Meta Quest 3 and the Android XR devices arriving from Samsung and others have raised what is possible: hand tracking, eye tracking, room-scale mapping and passthrough good enough to work in for an hour. They are also expensive, niche, and mostly used by developers and early adopters. For a product team that means two things.
First, do not design for headsets unless your users own them — and for almost every consumer product, they do not. Second, do design your content so that it can move to a headset later. A well-built 3D asset pipeline, sensible units, physically based materials and clean scene structure carry over. The phone experience you ship today is the foundation for the spatial one you may ship in three years, so it is worth building the assets properly the first time.
The engineering problems that actually decide the outcome
AR projects rarely fail because the tracking library was wrong. They fail on the unglamorous parts around it:
- Asset weight. A 60 MB model that looks superb on a workstation will never load on a phone over mobile data. Budget polygon counts and texture sizes from day one, and build an export pipeline that enforces them.
- Lighting and scale. An object that floats slightly above the floor, or is a few percent too large, breaks the illusion instantly. Real-world testing on the actual devices your customers use is not optional.
- Onboarding. Most people have never scanned a floor with their phone. The first ten seconds — "move your phone slowly to find a surface" — are where the majority of drop-off happens, and where good copy and a clear indicator earn their keep.
- Fallbacks. Some devices will not support the experience, and some users will deny camera permission. A 3D viewer without the camera, or a plain image gallery, keeps those people on the path instead of on an error screen.
- Measurement. Instrument the flow like any other funnel: opened, surface found, object placed, interacted, converted. Without those events you will never know whether the feature is working or merely present.
Choosing the right kind of tracking
"AR" covers several quite different techniques, and choosing the wrong one for the job is a common early mistake. Surface tracking finds floors and tables and lets a user place an object in the room; it is the right tool for furniture and appliances, and the one with the most demanding onboarding. Image tracking recognises a printed target — a label, a poster, a page in a manual — and anchors content to it; it is reliable, cheap, and ideal for packaging and print. Face and body tracking powers try-on for glasses, cosmetics and some apparel. Location-based AR uses GPS and, increasingly, visual positioning against mapped streets to anchor content outdoors, which is what powers walking directions and city-scale campaigns.
Each has a different accuracy, a different device requirement and a different failure mode. A guided-repair tool built on surface tracking will drift; the same tool built on image targets printed on the equipment will stay locked. Getting this decision right in the first week saves a rebuild in the third month.
How to scope a first AR project
Pick one moment of hesitation in your customer's journey — the point where they stop and think "I am not sure" — and build the smallest overlay that answers that specific doubt. Ship it on the web, put it in front of real users, and measure the behaviour after it: did they buy, did they finish the setup, did they call support less. Only then decide whether the moment deserves a native app, a richer experience, or a quiet retirement.
That discipline is what separates AR that earns budget from AR that gets demoed at a conference and never mentioned again. It keeps the scope small, the cost predictable, and the result something people actually use twice.
Frequently asked questions
Do we need a native app to offer AR?
Usually not for a first version. WebAR delivers product previews, guided instructions and interactive print through a link, with no install. A native app becomes worth it when you need more stable tracking, offline use, or deeper device integration — and by then you will have usage data to justify it.
How long does an AR feature take to build?
A focused WebAR product viewer with a handful of optimised 3D models is typically a few weeks of work, most of it in asset preparation and testing on real devices. Guided-instruction or industrial overlay tools take longer because the content and the tracking targets need to be built and validated for each scenario.
Is AR worth it for a small e-commerce store?
It depends on what you sell. For items where fit, scale or appearance in a space drives returns — furniture, lighting, décor, some fashion — an AR preview can pay for itself in reduced return handling alone. For small, low-consideration products it rarely moves the needle, and the budget is better spent elsewhere.
AR works best when it solves a specific moment of confusion or hesitation. We help teams find that moment, design the overlay around it, and build something people actually use twice.