Web Analytics

Before any tool goes in, there is a measurement plan: which actions count, what they are called, and what additional information travels with them. Without that list you end up with events under similar names that nobody can tell apart later.

Depending on the situation we implement it with Google Analytics 4, with low cookie tools such as Plausible, or with a self hosted Matomo instance. The data comes from a data layer in the source code, and for shops from the usual events along the purchase path: product view, add to cart, checkout start and order. For WooCommerce and Shopware we build that into the theme or through a plugin, so cart and order values come from the system rather than being scraped off the page.

Sources that need no consent come on top: server logs, status codes, loading time measurements and Search Console data. The result is a report with a manageable number of metrics, each one tied to a decision it supports. Everything else stays available inside the tool without having to occupy anyone every month.

Web analytics earns its keep when enough is happening for differences to be visible. In a shop with steady orders you can read off where in the purchase path customers drop out and whether a change had an effect. On a lead focused site with a steady flow of inquiries you can attribute which pages and sources lead to conversations.

Common triggers are an upcoming relaunch that needs a reliable baseline, a shop with a striking number of checkout abandonments, unclear origins of inquiries ahead of a budget decision, or an existing setup whose numbers nobody believes any more, because orders get counted twice or a traffic source suddenly disappears.

With very few visits and a handful of inquiries per month, the numbers carry no statement at all. Chance drowns out every effect and a test would have no solid basis. In that situation, customer conversations, recordings of individual sessions and a careful review of forms and checkout deliver more than any analytics tool. Simply counting inquiries is usually measurement enough at that stage.

Analytics tools that read information from a device or store identifiers generally require consent. In practice that means a consent tool that separates categories cleanly, fires tags only after agreement, and keeps a record of decisions. How your specific case is to be judged legally is a question for your data protection or legal advisors. We implement what they specify, document which service starts under which category, and then verify that a rejection really does stop everything.

Server side tracking through an endpoint on your own domain makes the transfer more stable and keeps data controllable, because you decide what gets passed on. It does not remove the need for consent, it simply moves where that question is examined.

When attributing inquiries, a gap always remains. Banner rejections, ad blockers, shortened cookie lifetimes in some browsers and switching between devices mean part of the journey stays invisible. So we compare several sources against each other: advertising accounts, the analytics tool, the CRM, and a short question in the form asking how someone found you.