Web Analytics Fundamentals

The Core Units of Measurement

User — a unique visitor, identified by a client ID stored in a cookie. The same person on phone and laptop counts as two users. Clearing cookies creates a new user.

Session — a group of interactions within a time window. One user can have many sessions. Sessions end after 30 minutes of inactivity by default.

Event — a single interaction: a page view, a click, a scroll, a purchase. In GA4 everything is an event, including page views.

The relationship matters when reading reports. A conversion rate calculated per session differs from one calculated per user, sometimes substantially.

What Analytics Cannot Tell You

Analytics data is systematically incomplete and it is important to know how:

• Ad blockers and privacy browsers prevent a meaningful share of tracking
• Cookie consent rejection removes users entirely from measurement
• Cross-device journeys fragment into separate users
• Bot traffic inflates figures unless filtered
• Dark social — links shared privately — arrives as direct traffic with no attribution

Treat analytics as a large, consistent sample rather than a census. Trends and relative comparisons are reliable; absolute totals are not.

Metrics That Matter vs Vanity Metrics

A metric is useful if a change in it would cause you to do something differently.

Pageviews rising tells you little on its own. Conversion rate by landing page, revenue per session by channel, or the drop-off rate at a specific checkout step all point at a decision.

The most common error is optimising a proxy metric until it decouples from the outcome it was meant to represent — driving low-quality traffic that improves session counts while revenue stays flat.

Building a Measurement Plan

Before configuring anything, write down the business objectives, the user actions that indicate progress toward them, and the metrics that quantify those actions. Only then decide what to track.

Sites instrumented without this step end up with hundreds of events nobody reads and no clear answer to whether the site is working. A short list of well-defined conversions beats exhaustive tracking.

Where the Numbers Come From, and Why Two Tools Disagree

Two analytics tools will never agree, and knowing why prevents a great deal of wasted reconciliation.

Different session definitions. How long a gap ends a session, whether a new campaign starts a new one, and what happens at midnight all differ between tools.

Different collection points. A tag firing on page load, a tag firing after consent, and a server log capture three different populations of the same traffic.

Different bot filtering. Each vendor removes automated traffic on its own rules, before you see any of it.

Different attribution. One tool's last-click and another's data-driven model will allocate the same conversion differently, and both are internally consistent.

The practical rule: pick one system of record per question and stop reconciling. Use the ad platform's numbers for optimising the ad platform, your analytics for on-site behaviour, and your own database for revenue. Trying to make three systems agree consumes analyst time and produces nothing, because they are measuring different things by design.

A Measurement Plan That Fits on One Page

Most analytics implementations begin with tools and end with a dashboard nobody uses. The order should be the reverse.

The one-page structure that works:

1. The business objective, in a sentence. Not grow — what specifically.
2. The two or three questions you need answered to steer towards it.
3. For each question, the metric that answers it, and the comparison that makes it meaningful — against last period, a target, or a segment.
4. For each metric, where it comes from and who owns it.
5. What you will do differently depending on what it says.

Step five is the one that gets skipped and the one that matters. A metric with no decision attached is a number you will report monthly for two years and never act on — the same test that governs a segmentation, a dashboard and a clean room.

Write it before configuring anything. The plan determines what you need to track; the tool does not determine what you need to know.

Sources

What each claim on this page rests on. Entries are typed so you can see which are primary.