Analytics that matter

Start from the decision

Before looking at any number, ask what you would do differently depending on the answer. If nothing, the number is not worth your time however interesting it is.

That single question removes most of a standard analytics dashboard. Sessions, bounce rate and average time on page rarely change anything for a business of this size.

It also makes the remaining numbers useful, because each one is attached to a decision you had already identified as worth making.

It is also a defence against dashboards. A tool will happily show you sixty numbers, and the four you chose deliberately are worth more than all of them together.

Write the decisions down first, in a sentence each. Then find the number that answers each one, and ignore everything the tool shows you that is not on that short list.

The four that earn a look

Enquiries, first and above everything. How many people contacted you through the site, and whether that is going up or down, is the number the site exists to produce.

Then where visitors came from, which pages they landed on, and what they searched. Those three explain the first one and point at what to do next.

Our page on the search performance report covers the last of them, which is the one most small businesses never open.

Those four fit on a single page and can be read in a few minutes. That is the point: a set small enough to actually look at is a set you will actually look at.

What to actually track

The list below is a complete measurement setup for a small business site. It takes an afternoon and then answers questions rather than generating reports.

Anything absent from this list is not forbidden, it is simply not worth the monthly minutes on a site of this size. Add something only when you can name the decision it would change.

• Form submissions, counted reliably
• Phone taps on a mobile device
• Which pages people arrive on
• Where they came from, by source
• What they searched to reach you
• Which pages nobody reaches at all
• Nothing that would not change a decision

Count enquiries properly

The most common measurement fault is not knowing how many enquiries the site produced. Analytics counts visits, and visits are not the outcome.

A confirmation page at its own address is the simplest reliable count. Our page on thank you pages covers why that page is worth having anyway.

Count phone taps too. On a mobile site a large share of enquiries never touch a form, and a business measuring only form submissions undercounts itself badly.

Then compare that count against what you know. If the site says forty enquiries and you remember twelve, something between the form and your inbox is failing and needs finding.

Privacy is part of the setup

Analytics tools set cookies and send visitor data to a third party, which carries obligations depending on where your visitors are.

Our page on analytics and privacy covers the choices, including lighter tools that answer the same four questions without personal data.

For a small business that is frequently the better trade. Fewer numbers, no consent banner, and the questions that mattered still answered.

Whichever you choose, say so on the page. A short, genuinely useful privacy note explaining what is measured costs nothing and is the kind of detail people notice when it is missing.

Look quarterly, not daily

Small site numbers are noisy. A week is too short a period to mean anything, and daily checking produces reactions to variation rather than to change.

A quarterly look at the four numbers, compared with the same quarter last year, shows what is genuinely moving. That is a twenty-minute job.

Write down what you concluded. A note beats a saved report, because the conclusion is the part you will want in six months and the report will still be there anyway.

Then do one thing as a result. A quarter that produces a conclusion and no change was a quarter of reading, and reading is not what the measurement was set up for.