AR navigation · 2024
Live View
Turning a phone camera into a wayfinding instrument — and figuring out how much of the world a person needs to see before they trust the arrow.
Defining the problem
People don't get lost on the map. They get lost in the first thirty seconds.
Placeholder — please confirm the details. We watched people leave a station exit and stand still, rotating the phone until the blue dot agreed with them. The failure wasn't routing; it was orientation. Everything after that moment worked fine.
That reframed the brief: instead of a better map, we needed a faster way to answer one question — which way am I facing right now?
Process
Eleven street tests across three cities, each one a working build rather than a mockup — camera behaviour is impossible to judge from a static frame.
Anchor first
Confirm the world before drawing anything into it.
Show the seam
When tracking is uncertain, say so — silence reads as a lie.
Hand it back
Camera up is a posture, not a mode. Get people back to the map fast.
What we learned
Trust is built in the first two seconds, and it is spent every time the arrow lies.
Placeholder outcome copy — happy to swap in the real numbers. The version that admitted uncertainty was rated more accurate than the version that hid it, even though the tracking was identical.