Five years of eye tracking in user research.

We wrote about our eye-tracking lab in early 2020. Five years on, the gear is better, the methodology is sharper, and the patterns we keep finding have surprising consistency across very different product categories. Some reflections on what changed and what didn't.

What the gear looks like now

• 01Tobii Pro Spectrum has been replaced by the Tobii Pro Fusion — same data quality, smaller, more portable.
• 02Tobii Glasses 3 for in-context retail and installation testing. Wearable, head-mounted, real-environment tracking.
• 03Software (Tobii Pro Lab) finally has a usable cross-session aggregation view — saves hours of manual gaze-plot synthesis.
• 04Remote tracking — webcam-based, lower quality, mostly useful for smoke-tests rather than serious research.

What we know now that we didn't in 2020

• 01Time-on-page is a worse metric than attention duration. Users can be on a page and not looking at it.
• 02Heatmaps oversell. Gaze plots and AOI dwell times are where the real signal is.
• 03Six users still finds 80% of the major issues. We've tested with twelve and the marginal returns are real but small.
• 04Mobile eye-tracking is a different sport — we use a chin rest and a fixed test rig for it.
• 05Cognitive load shows up in pupil dilation. We don't measure it routinely, but when we have it's pointed us at the right friction every time.

Patterns that keep showing up

• 01Headlines get fixated, then the eye drops below for the supporting copy — but only sometimes reads it. Make the headline carry the message.
• 02Visual hierarchy drives gaze order more than reading order. Designers who set type sizes properly are doing UX work, even if they don't realise it.
• 03Forms get scanned to find the end before they get filled. Long forms abandon at the same rate regardless of complexity.
• 04Pricing tables are read right-to-left almost as often as left-to-right. The right-most column being 'best for most teams' performs unusually well.
• 05Hero CTAs are noticed when they look like buttons. Otherwise they get treated as decoration.

"

There's no single tool we've added to the studio that's changed more product decisions per session than the eye-tracking rig has.

What we'd tell anyone setting up a lab today

• 01Buy the mid-tier hardware. The flagship is overkill for almost every brief.
• 02Invest in the AOI tooling. Defining good Areas of Interest is what turns raw gaze data into product decisions.
• 03Use the lab early — pre-launch, even pre-prototype. The most valuable sessions are the ones that change scope, not the ones that confirm it.
• 04Pair it with task analytics in production. Lab data tells you why; product analytics tells you whether the why is the why for most users.

The lab still earns its place every quarter. There's no single tool we've added to the studio that's changed more product decisions per session than it has.