As an analytics guy, we deserve some blame, but also can and should be part of the solution. Anybody working on a public website in the last 12 months has had t

As an analytics guy, we deserve some blame, but also can and should be part of the solution. Anybody working on a public website in the last 12 months has had their senses sharpened about page speed issues thanks to Google’s carrot of AMP-free news placement for fast sites and a badge of shame and lower rankings for slow sites. Lighthouse has been instructive. Apps across platforms from mobile, desktop, smart TV, auto, and more could benefit from the same treatment. - Analytics packages need to shrink. Most can be under 5kb, but rarely are. - excessive event tracking and demanding items like scroll listeners need to die - Ad tracking is completely nonchalant about speed, and oftentimes chains dozens of tags (or hundreds to thousands for header bidding) Analysts can save the day by tracking the right metrics: - time and clicks in navigation should be tracked and optimized for minimal use. Get users to content fast. - rage buttons. No lengthy surveys, just allow users to indicate they are angry at a feature. Count the times it gets clicked per location. - Optimize towards retention signals first. They matter far more than conversions. Look at the usage patterns of users that don’t return and/or churn. The top-line funnel will sing if users are pleased with the UX. This matters as much for desktop tools as it does for SaaS. - Each deployment with UX implications deserves tagging and long-term measurements. Look for how long it takes users to adapt to changes, and not just when it’s brand new. More clicks is probably worse, don’t brag about that. - Sometimes a longer session means more contented usage, but often it means more frustration. - pay close attention to engagement with settings UX. That’s where people go when the UX isn’t right. - Give users some access to the dashboards you look at, if feasible. I’m a firm believer that people being measured should be able to observe some form of what is derived from that data. If PMs exist to improve user engagement, engage them in the process. Two-way mirrors, whether virtual or physical, are creepy AF.