Non-confusing UIs for analytics
· 3 min read
Open a typical analytics product and count the numbers on the first screen. Forty is common. Sixty is not unusual. Sessions, engaged sessions, engagement rate, average engagement time, events per session, conversions, new users versus returning users, and a dozen cards you have never clicked. The dashboard is not lying to you. It is just answering questions nobody asked.
An analytics UI has one job: let a person who did not build the tool answer "how is my site doing?" and "what changed?" without training. Everything else is optional. Here is how we think about it.
Show what people act on
Ask a site owner what they check and the list is short. How many people came. Which pages they read. Where they came from. Whether that is more or less than last week. Those four things drive nearly every decision a small team makes: what to write next, which channel to invest in, whether the launch worked.
So they go first, large, and above the fold. Visitors and pageviews as two big numbers. A chart of the last thirty days. Top pages and top referrers as two short lists. That screen is the product. Anything added to it has to earn its place by changing a decision.
Hide what people argue about
Bounce rate, session duration, and engagement rate are the metrics people spend the most time discussing and the least time acting on. They are also the least trustworthy. Session duration cannot be measured for the last page of a visit. Bounce rate depends on how you define a session. Engagement rate is whatever the vendor decided it was this quarter.
We leave them out. Not hidden in a settings menu, out. If a number needs a paragraph of methodology to be interpreted correctly, it is not helping a busy person. It is generating meetings.
One time range, one place
Every number on the screen should describe the same period, and the control for that period should be in exactly one spot. Dashboards that let each widget carry its own range are a trap: the visitor tile says thirty days, the chart says seven, and the referrer list quietly says "all time". People compare them anyway and draw conclusions from noise.
Lists beat charts for categories
Pie charts of browsers, donut charts of countries, stacked bars of devices: all of them are harder to read than a sorted list with a count. A list answers "which is biggest?" and "how big?" at a glance. A pie chart answers "which is biggest?" only when one slice dominates, and never answers "how big?" without a legend and some squinting.
Use charts for time. Use lists for everything else. A small bar behind each row in the list gives the visual sense of proportion without the reading problems.
Name things what they are
"Visitors" means people, counted once per day. "Pageviews" means page loads. "Referrers" means the site the visitor came from. These words are good. "Users", "sessions", "acquisition", and "engagement" each need a definition and the definitions move.
Pick plain words and keep them stable. If a metric changes meaning, rename it.
Make empty states useful
The first screen a new customer sees is empty. That screen should tell them exactly what to do, with the snippet ready to copy, and should turn into the real dashboard the moment the first pageview arrives. An empty chart with a "no data" label is a dead end. A page that says "paste this, then reload" is an onboarding flow.
Make the loud thing quiet
Notifications, insight cards, "anomaly detected" banners, AI summaries. They all interrupt the reading of the actual numbers. If the chart is clear, the person will see the spike. If the chart is not clear, no banner will fix it.
Design for the second look
The first time someone opens a dashboard they explore. Every time after that they are checking. Checking should take five seconds: open, glance at two numbers and one chart, close. Layout stability matters more than density. Numbers should sit in the same place tomorrow as today. Nothing should shift when data loads.
The rule behind the rules
Whenever we consider adding something to the dashboard we ask: would a customer make a different decision because of it? If the honest answer is "they would find it interesting", it does not go in. Interesting is cheap. Clear is the product.