Reading website analytics properly: the five numbers that actually matter
Most analytics tools give you more metrics than anyone ever opens. Here is which ones describe your website honestly, which are noise, and how to read them without fooling yourself.
July 14, 2026 11 min read
Most dashboards measure the wrong thing first
Open a general-purpose analytics tool and the first number you see is usually sessions, followed by a bounce rate defined differently by every vendor who reports it. Neither tells you what to change. Pageviews, unique visitors, pages per visit, top pages and referrers do, because each one maps directly onto a decision you can actually make about your site.
The discipline is not collecting more. It is refusing to display a metric you cannot act on. A number that never changes a decision is a number that costs you attention every time you look at it.
Pageviews and visitors answer different questions
A pageview counts what was read. A unique visitor counts who read it. When the two move together, your traffic is growing. When pageviews rise and visitors do not, the same people are reading more — often the sign of a page that finally earned its place in your navigation.
In Lomen, visitors are derived without cookies from a short-lived, non-reversible grouping of the request. That is slightly less precise than a cookie-based count and considerably more honest about what it knows. It is enough to count a visit, and not enough to recognise a person a week later.
Referrers are where spikes get explained
Almost every unexplained jump in a traffic graph has a referrer behind it. A link landed somewhere, brought a burst of attention and then faded. If you only look at totals, you will notice the spike after the attention has already gone.
The useful habit is to read referrers and entry pages together. A referrer tells you where people came from; the entry page tells you what they came for. Once you know both, you know whether the page they landed on deserved them.
Device splits are a layout signal, not a demographic one
Device class is worth measuring for exactly one reason: deciding which layout deserves your next afternoon. If mobile is a third of your visitors but a tenth of the views on the page that matters most, the problem is rarely intent. It is usually load time or a layout that buries the thing people came for.
That is also why Lomen collects device class and nothing more. No fingerprinting signals, no device identifiers, no cross-site profile — none of it would improve the decision the number exists to support.
A review routine that takes ten minutes
Once a week, look at pageviews against the previous period, then the top five pages, then the top five referrers. Write down anything that moved more than your usual weekly variance. That is the whole routine, and it will catch nearly everything worth catching on a site of any normal size.
This is precisely the routine Lomen is built around, but the routine matters more than the tool. Whether you run it by hand or let something run it for you, the point is that traffic truth is a weekly read, not a number you glance at between other work.
Lomen runs this weekly read for you
Pageviews against the previous period, top pages, referrers and device mix on one screen — from a single line of script, with no cookies and no IP storage.
Add your first site on Basic