Choosing which website metrics to track

Decide what the site is for before you measure anything

A website usually has one primary job and two or three secondary ones. For most of the businesses we work with the primary job is producing qualified enquiries. For a festival it might be ticket sales. For a research centre it might be attracting the right postgraduate applicants.

Whatever it is, write it down in a sentence, and get whoever runs the business to agree with it. That sentence determines everything else, and it's remarkable how often two people in the same company would write a different one.

Without it you end up tracking whatever the tool shows by default, which is traffic, and traffic is the least useful number available to you.

The numbers worth watching

Enquiries, obviously, but broken down by where they came from and whether they were any good. Twenty enquiries from people who can't afford you is worse than four from people who can, and a raw count hides that entirely.

The route people take to the enquiry form. Which pages they read first, where they drop out, how many come back later. This is where you find the actual problems, and it's the part most reports leave out.

How your key pages perform against each other. If your services page converts at three times the rate of your homepage, that tells you something worth acting on.

And a small number of specific searches you want to be found for, tracked over months rather than weeks.

The numbers that mislead

Sessions and pageviews on their own tell you almost nothing about whether the site is working. They go up when you run ads, when a competitor mentions you, and when a bot has a busy week.

Bounce rate is routinely misread. Someone landing on your contact page, finding the phone number and leaving is recorded as a bounce, and that's a completed job rather than a failure.

Time on page is similar. Longer isn't better if people are spending it confused. We've seen a page's average time drop sharply after a rewrite while enquiries from it doubled.

Set it up so it answers questions

Track the things that represent progress towards the job you wrote down. Form submissions, phone clicks, booking clicks, downloads of the thing people download before they buy.

Tag them well enough that you can tell them apart six months later. A single event called submit is not much use when you have four forms.

Record the source properly so you can answer where enquiries actually come from, which is the question every owner asks and few analytics setups can answer.

None of this is difficult and most of it takes an afternoon. It's simply skipped, usually because the site launched and everyone moved on.

Review it in a way that leads somewhere

Once a quarter is enough for most businesses. Monthly reviews on low-traffic sites mostly measure noise, and weekly ones guarantee that nobody looks properly.

Have the review produce a decision, even a small one. One page to rewrite, one form to shorten, one thing to test. A review that ends without a decision is a report, and reports don't change anything.

Write down what you expected to happen before you make the change. Otherwise you'll rationalise whatever result you get, which everyone does and nobody notices.

A reasonable starting point

If you're beginning from nothing: install a privacy-friendly analytics tool, set up conversion tracking on your enquiry form and phone links, connect Search Console, and agree the one sentence about what the site is for.

That's perhaps half a day of work and it puts you ahead of most businesses of your size. Everything more sophisticated can wait until you've run a couple of quarterly reviews and know which questions you actually keep asking.

The goal isn't a comprehensive dashboard. It's being able to answer, at the end of a year, whether the money you spent on the website did anything.