Web analytics

Web analytics answers one question in many variations: what do people do on your website, and what stops them doing the thing the website exists for? Tool selection, event definitions and reports are means to that end, not the end itself.

This page describes how measurement is set up today: what data model current tools work on, which metrics say something, how consent is wired in technically, and when the effort of server-side tagging pays off.

The data model has changed

Anyone still looking for goals, bounce rate in the old sense, or sessions as Universal Analytics defined them is looking in vain. Universal Analytics was switched off in July 2023, and a year later processing of the historical data ended too. Guides and reports still built on that structure lead nowhere.

Google Analytics 4 does not measure page views with additional information, it measures events. A page view is an event, a click is an event, a submitted basket likewise. Every event carries parameters, and the analysis emerges from how those parameters are combined. That is more flexible than the old model and considerably less forgiving: whatever was not defined as an event appears in no report.

What used to be called “goals” are now key events – the events you explicitly mark as important. In Google’s world the term conversion is now reserved for the advertising products, which regularly causes confusion in reports as soon as Analytics and Google Ads sit side by side.

Engagement rate has taken the place of bounce rate. A session counts as engaged if it lasts longer than ten seconds, covers at least two pages, or triggers a key event. In GA4 the bounce rate is now merely the inverse of that.

Consent is part of the measurement, not an appendage

Tracking scripts may only load once the visitor has actively agreed. A pre-ticked box, or a banner showing only an “accept” button, does not meet that requirement. Technically this means the consent management tool sits in front of all measurement scripts and releases them individually, instead of letting them run in parallel and cleaning up afterwards.

Anyone using Google advertising products in the European Economic Area has had no way around Consent Mode v2 since March 2024. It supplements the familiar signals for analytics and ad storage with two further parameters recording whether user data may be transmitted to Google and used for personalisation. Without those signals, audience lists and conversion import into Google Ads simply stop working.

The practical effect of the consent requirement: a portion of your visitors is not measured. Google Analytics 4 fills that gap through modelling, that is, through estimates derived from the measured share. That is usable for trends and useless for statements about individual sessions. Confuse the two and you make decisions on a number that never existed.

Cookieless and privacy-friendly alternatives

Google Analytics is widespread, but it is not the only option. For websites where the effort around consent and third-country transfer is out of proportion to the insight gained, other tools are often the better choice.

Matomo is open source and can be run on your own servers. The data then never leaves your infrastructure. In its cookieless configuration Matomo does without recognition via device storage, which makes the numbers somewhat coarser and simplifies the legal construction considerably.

Plausible takes a deliberately reduced approach: few metrics, one report, no cookie, EU hosting. For company websites and blogs that is often enough. For a shop with a multi-step checkout and product list analysis it is not.

Piwik PRO sits in between and brings consent management, tag management and analytics together in one package, hosted in the EU.

The honest trade-off: Google Analytics 4 has the deepest integration with Google’s advertising products and the largest body of documentation. Everything else is simpler to operate, easier to justify, and delivers less data. Which side wins depends on whether you want to steer ads or improve content.

Server-side tagging

In the classic setup, the visitor’s browser loads the scripts of every service involved and sends data directly to their servers. With server-side tagging, a container of your own runs under a subdomain of your website. The browser only talks to that container, which decides what it passes on to which service.

That delivers three concrete things. First, less third-party code in the browser, which noticeably eases loading time. Second, the ability to trim data before passing it on – IP addresses, say, or parameters nobody outside needs. Third, more robust measurement, because ad blockers and browser restrictions hit your own endpoint less often.

What server-side tagging explicitly is not: a way around consent. The obligation attaches to access to the device and to the processing of personal data, not to where the container sits. Running it also costs money and attention; for small websites it rarely pays off.

Metrics that say something

Most reporting fails not from missing data but from too much of it. The useful metrics are the ones an action follows from.

• Key events per channel show which route actually leads to enquiries or orders, rather than merely delivering visits.
• Drop-outs along a defined funnel name the point where people leave. In a checkout that is often a single step, not the process as a whole.
• Engagement per content type separates pages that get read from pages that merely get opened.
• Internal search is underrated: search terms with no results are a list of what visitors expect and do not find.
• Click-through rate connects visibility with actual interest.

For checking immediately after a change, real-time reports are the right tool – not for analysis, but to see whether a newly configured event arrives at all.

From measurement to decision

Clean measurement is the precondition, not the result. The value appears when an anomaly turns into a hypothesis and the hypothesis into a tested change.

For shops a second layer comes in: basket value, return rate and repeat purchase rate appear in no standard report side by side, yet they determine whether a channel pays for itself.

What we take on

• Setting up or reworking the measurement, including event definitions and data layers
• Selecting and configuring the tool, including outside Google Analytics
• Integrating a consent management tool and verifying that scripts really do load only after consent
• Setting up and running server-side tagging where it pays off
• Migrating historical data and building comparable reports
• Regular reporting on the metrics that matter for your business
• Training, so that your team reads and interprets the reports themselves

If you are unsure whether your measurement is correct, a review of the existing setup is the most sensible first step: we look at what is measured, what is missing and what is counted twice. Tell us your starting point – you can also book a slot directly there.