Projects Blog Contact
12 January 2023

2 min read

Google Analytics: what to configure on day one

Almost every website has Analytics installed. Very few have it measuring anything useful. This is what I always check, in this order.

1. That it's your property

It sounds absurd, but it's the most expensive fault I've come across: a measurement ID copied from another site during a migration. Months of data going to a property that wasn't theirs, and their own property sitting empty with nobody noticing.

Check it today: look at the code on your site, look at the ID of your property in Analytics, and confirm they're the same.

2. That it respects consent

If you have a cookie banner, check Analytics doesn't load until someone accepts. If the banner appears but the script has already loaded, the banner is decorative and you're not compliant.

3. Events, not just page views

GA4 measures some things on its own (scroll, outbound clicks, downloads), but the important ones you have to define:

And mark them as conversions. Without this, Analytics tells you how many people passed through, not how many did something.

4. Exclude your own traffic

Filter out your IP and your team's. If you work on the site daily, your visits pollute every report, especially on a low-traffic site.

5. Search Console connected

It's free and brings search queries into Analytics. It's the difference between knowing people came from Google and knowing what they were looking for.

6. Reports you'll actually open

Three reports are enough for a small business: acquisition (where they come from), pages (what they read) and conversions (what they do). Everything else is curiosity.

7. Cross-check against another source

Analytics always sees less than what happens: blockers, people declining cookies, browsers with protection. On one of my own high-traffic projects, GA4 only saw between 30% and 36% of the visitors the server counted, and that ratio held even on the peak day.

The practical conclusion: for reach, use server or CDN data; for behaviour, use Analytics. And if you quote a figure to anyone, say where it comes from.

Want me to check whether your analytics measures what you think?

Write to [email protected]. It takes half an hour and there are usually surprises.

← Back to the blog