Attach an analytics tool and the screen fills up fast: sessions, pageviews, average time on page, bounce rate, traffic sources. Watching numbers grow feels productive, yet look back a few months later and it is common to find that not one of those numbers changed a single decision. The dashboard grew; the decisions did not.
This series starts with deciding what to measure rather than with tool setup. Do it the other way round and you simply inherit whatever the tool reports by default — and that list reflects what is easy to collect, not what your business needs to know.
There is really only one test
Before adding a metric, ask yourself one sentence: “if this number moved noticeably, what would I do differently?” A concrete answer means the metric is worth managing. An answer along the lines of “I suppose I would be pleased” means it is a background number, and background numbers do not belong at the centre of a weekly review.
That question alone tends to halve the list. The shorter list then gets used more often — five things get checked every week, forty things get opened by nobody.
Vanity numbers and decision numbers
Pageviews rising by a fifth tells you nothing on its own. Whether that traffic turned into enquiries or arrived on an irrelevant query and left immediately points to opposite responses. By contrast, if the share of people who open the form and go on to send it falls, you have an immediate reason to look at the form.
None of which makes the left column useless. Those numbers are a reasonable way to sense scale. They simply belong in the background, where background numbers live.
Three things a candidate metric must pass
The third one fails most often. When the definition changes halfway through, the before and after become two numbers you cannot compare. So write a one-line definition beside every metric — what exactly is being counted — and, if you ever revise it, record that you did.
Four is enough to start
The bottom layer may look out of place. But a meeting held on figures inflated by office traffic and bots produces decisions that are wrong from the start, so quality is not one metric among others — it is the precondition for all of them. The final part is devoted to it.
More on designing search traffic and measurement together sits in the SEO archive, and if you would like the site’s own speed and indexing reviewed at the same time, a diagnostic is part of the optimization program.
Next part
With the metrics settled, the tooling comes next. The following part compares the three ways of putting analytics into WordPress — and the price each one quietly charges.