The GA4 mental model: everything is an event

Start with a question, not a report

Google Analytics 4 is a system for recording what people do on a website or app. Its real value is not the number of charts it can make. Its value is helping you answer a useful question: Where do new customers come from? Which article earns newsletter sign-ups? Where do shoppers leave the checkout?

Begin with the decision you want to improve. Then choose the action that gives you evidence. If your question is “Is our new landing page working?”, a page view only tells you that the page loaded. A sign-up, purchase, or qualified lead tells you whether something valuable happened next.

The four building blocks

A user is a person or device interacting with your site or app, as estimated by the identifiers available to GA4. A session is one visit within a period of activity. An event is a distinct interaction, such as a page view, click, sign-up, or purchase. A parameter adds detail to an event, such as the product name, button location, or purchase value.

One user can have many sessions. One session can contain many events. One event can carry many parameters. Keeping these levels separate helps you avoid comparing numbers that describe different things.

• User: who is interacting, within the limits of available identifiers.
• Session: the visit in which interactions happen.
• Event: what happened.
• Parameter: useful detail about what happened.

Read an event like a sentence

An event name is the verb. Parameters are the extra words that make the sentence specific. “add_to_cart” tells you the action. Parameters such as item_id, item_name, currency, and value tell you what was added and what it was worth.

GA4 uses this event-based model across websites and apps. That makes it flexible, but it also means your setup needs thought. A clear event name and a small set of useful parameters are usually more valuable than dozens of vague events.

event: add_to_cart
parameters:
  item_name: "Everyday Sneaker"
  item_id: "SKU-104"
  currency: "USD"
  value: 79

Put users, sessions, events, and items on the right level

Most confusing GA4 tables become clearer when you ask, “What does one row represent?” A row can represent a user, a session, an event, a page, a product, or a campaign. Those are not interchangeable units. If three product views happen in one session, the session count is one while the event count is three.

Scope is the level at which a value applies. A campaign that first acquired a user has user scope. A campaign that started the current visit has session scope. A button name belongs to one event. A product brand belongs to an item. When you combine fields, check that their scopes tell a coherent story.

This matters for ratios too. Key events per session is not the same as key events per user. One person can return twice and convert once, producing a different answer depending on the denominator. Write the denominator into your analysis notes before comparing rates.

Learn the small set of metrics you will see every day

Views counts page or screen views. Event count includes all recorded events, so it will usually be much larger than views. Sessions counts visits, while active users estimates the people or devices behind meaningful activity. None of these is inherently “the best” metric; each answers a different question.

Engaged sessions and engagement rate attempt to separate meaningful visits from quick, low-value activity. They are useful context, but they are not a substitute for your business outcome. A long visit can be confusion rather than interest, and a short visit can still answer someone’s question successfully.

Key-event rate shows the share of users or sessions associated with a key event, depending on the report context. Always read the metric definition in the interface and record the exact name. “Conversion rate” in a stakeholder slide is ambiguous unless you explain the numerator, denominator, and key event involved.

Follow one complete journey from arrival to outcome

Imagine Maya sees a newsletter, opens a product page, returns the next day through search, adds a shoe to her cart, and buys it. The newsletter may receive first-user acquisition credit, search may receive session acquisition credit for the return visit, and the purchase event carries transaction and item details.

During the purchase session, GA4 might record session_start, page_view, view_item, add_to_cart, begin_checkout, and purchase. Those events form a measurable sequence. Parameters explain which product was viewed, which campaign started the visit, and how much revenue the order represented.

An analyst would not stop at “purchase increased.” They would compare sessions, purchasers, completion rates, products, devices, and acquisition paths. They would also check whether the purchase tag changed, whether a promotion altered demand, and whether consent or tracking gaps affected the visible population.

Build your first measurement question canvas

A question canvas is a short bridge between a business request and an analytics build. Write the decision, the population, the action, the context, and the expected use. For example: “The growth team wants to know whether the spring email campaign generates demo requests from new visitors on mobile.”

This sentence already suggests a campaign dimension, a new-versus-returning comparison, a device breakdown, and a generate_lead event. It also suggests what not to optimize for: raw email clicks without a downstream outcome. If the team only needs visits, the event design can be simpler; if it needs lead quality, CRM or offline outcome measurement may also be required.

Finish with a feasibility check. Can the website reliably detect the action? Will the campaign links be tagged consistently? Is the population large enough to interpret? Are there privacy constraints? A few minutes spent here prevents hours of attractive but irrelevant reporting.

Worked example: measure a nonprofit resource journey

Imagine a nonprofit publishes a free analytics checklist. A visitor arrives from a partner newsletter, reads the guide, downloads the checklist, and leaves. They return three days later through organic search and make a donation. The organization wants to understand both content usefulness and donation behavior.

The first visit contains session_start, page_view, scroll if enhanced measurement records it, and a custom or recommended download event appropriate to the implementation. The second contains a new session_start, page_view, begin_donation if the product supports it, and purchase or another carefully defined donation event. The download and donation are different business actions and should not share one vague “conversion” event.

A user analysis can connect both visits within identity limitations. A session analysis can compare the checklist visit with the donation visit. An event analysis can count downloads and donations independently. A first-user acquisition report can examine the newsletter’s acquisition role, while session acquisition examines the search return. Each view is useful because the question and scope are explicit.

Mental-model mistakes that distort later analysis

The first mistake is using “user” as if it always means one known person. The second is assuming a session is continuous proof of attention. The third is treating event count as unique actions by unique people. These are easy shortcuts when a stakeholder asks for a quick answer, but they can change the conclusion.

Another common error is adding parameters without deciding their scope or analytical use. More data does not automatically create more understanding. A small model with clear users, sessions, events, parameters, and items is stronger than a large model full of inconsistent values.

• Do not add users and sessions as if they measure the same unit.
• Do not call every page view an engaged reader.
• Do not divide key events by an unnamed denominator.
• Do not mix first-user and session acquisition in one unlabeled chart.
• Do not use item fields when you mean event parameters, or vice versa.
• Do not infer motivation from an event without additional evidence.