What to measure when your website is not a shop

Online shops have it easy. Someone buys something, the money arrives, and you can tell whether the website worked. Everyone else gets an analytics dashboard full of numbers that go up and down for no reason anyone can explain.

If your site’s job is to generate enquiries, applications, donations, bookings or sign-ups, here is how to build a set of numbers that tells you something true.

Name the action first

Before you look at any tool, finish this sentence: this website is working when a visitor does this.

Usually it is one of a small set: submits the contact form, calls the number, books a consultation, applies for something, donates, subscribes, downloads a document, registers for an event. Pick the one that matters most and one or two secondary ones. Three is plenty. If you name eight, you will track none of them properly.

This step is not analytics work. It is a decision about what the organisation is trying to achieve, and it belongs to you rather than to whoever configures the tracking.

The numbers worth watching

• Completed actions per month. The count of the thing you just named. This is the headline number and everything else exists to explain it.
• Conversion rate. Completed actions divided by visitors. This is the one that tells you whether the site is getting better, because it strips out the effect of traffic going up or down.
• Actions by source. Search, social, email, ads, direct, referral. Two hundred visitors from a newsletter who send four enquiries beat two thousand from social who send none.
• Cost per action, where you are paying for traffic. Spend divided by completed actions. This is the number that decides whether a campaign continues.
• Form abandonment. How many people start the form and how many finish it. A large gap is a design problem you can usually fix in an afternoon.
• Entry pages that lead to actions. Which pages people arrive on before converting. It is frequently not the home page, and that changes where you spend effort.

The numbers that will mislead you

Not because they are fake, but because they answer a question you did not ask.

• Total page views. Rises when a competitor’s staff read your site, when a bot crawls you, when an unrelated article gets shared. Fine as background. Useless as a target.
• Time on page. Long can mean engrossed. It can equally mean confused, or a tab left open over lunch.
• Bounce rate in isolation. A visitor who lands on your contact page, reads the phone number and leaves has bounced, and has also done exactly what you wanted.
• Followers and impressions. Easy to grow, easy to buy, weakly connected to anything. We focus on business metrics, not vanity metrics, and this is where that line earns its keep.
• Keyword rankings alone. Position one for a phrase nobody searches is not an achievement.

Track what happens off the website too

For most service organisations and nonprofits, the website hands over to a human. What happens next is where the value actually appears, and analytics cannot see it.

So keep a simple record outside the tool. Every enquiry: the date, roughly where it came from, whether it was relevant, and whether it turned into anything. A spreadsheet is fine. Over a few months this tells you something no dashboard will, which is whether the enquiries you are generating are the ones you want.

Thirty enquiries a month is a bad result if twenty-five are irrelevant and the volume is eating your team’s mornings. Quality of lead is a real metric and it is almost always collected by hand.

Make sure the tracking is honest

Broken measurement is worse than none, because it produces confident wrong decisions. Things to verify:

• Submit a real test enquiry and confirm it registers as a completed action.
• Exclude your own team’s traffic. A small site with five staff browsing it daily has badly distorted figures.
• Check that phone clicks on mobile are counted, if calls matter to you.
• Tag your campaign links properly, or everything from your newsletter will be filed as direct traffic.
• Confirm your cookie consent setup is not silently discarding most of your data.

Look at it less often than you want to

Weekly numbers on a small site are mostly noise. Twelve enquiries this week against nine last week is not a trend, and chasing it leads to changing things too quickly to learn anything.

Review monthly. Compare against the same month last year where you can, because most organisations have a seasonal shape they underestimate. Change one significant thing at a time and give it long enough to show an effect.

A one-page report is enough

Completed actions this month. Conversion rate. Where they came from. Cost per action if you are spending. One sentence on what changed and one on what you will try next.

If your current reporting runs to fifteen pages and nobody has read it since March, that is a sign the measurement was never tied to a decision. Start again from the sentence at the top of this article.