GA4 setup: the events worth measuring, and the ones that are not

GA4’s biggest problem is not missing data but too much of it. Out of the box it collects dozens of events, most of which change no decision. The dashboard looks full and nobody acts on it.

A useful setup starts by making the list shorter.

Answer this question first

“If this number moves, what will I do?”

Remove every metric with no answer. Session duration went up — what will you do? Pageviews fell — what action follows? If there is no answer, the metric is decoration.

You are usually left with three to five numbers. That is a good dashboard.

Events worth measuring

For a site selling services:

Event	Why it matters
form_submit (qualified)	The real conversion, with spam and tests stripped out
phone_click	A serious intent signal on mobile
email_click	High-intent users who skip the form
pricing_view	The strongest intermediate signal of purchase intent
case_study_view	Shows the visitor is in evaluation
scroll_90 (long content only)	Whether the content is actually read

In e-commerce these are replaced by the standard commerce events: view_item, add_to_cart, begin_checkout, purchase. Send them with the parameters GA4 expects, or the revenue reports stay empty.

Not worth measuring

• Automatic scroll events. GA4 counts 90% scroll by default, on every page. On a short contact page that tells you nothing.
• Bounce rate. In GA4 this derives from “non-engaged sessions” and cannot be read in isolation.
• Average session duration. High can be good or bad: the user is reading, or cannot find what they came for.
• Pages per user. Same problem.

Do not delete them — just take them off the decision dashboard.

Setting up key events

GA4 now calls conversions key events. Three rules:

1 · Keep the count low. Mark everything as a conversion and nothing is one. Three or four is enough.

2 · Measure the qualified version. Not raw form submissions, but enquiries that passed spam filtering and can actually be worked. Do that check server-side and fire the event from there.

3 · Assign a value. Pass an estimated value with each enquiry. Without it GA4 cannot compare campaigns — it counts three cheap enquiries the same as one large one.

Do not lose the source

This is the most common and most expensive mistake. A user arrives from an ad, fills in the form, lands in the CRM — and the CRM does not record which campaign sent them. The result: nobody knows which ad produced a customer.

The chain you need:

1. Capture utm_* parameters and gclid on the landing page.
2. Write them into hidden fields on the form.
3. Carry those fields into the CRM record on submission.
4. When the deal closes, write it back to the ad platform as an offline conversion.

Skip step four and the ad algorithm keeps optimising for form count. It will find you the cheapest form — not the most profitable customer.

Three common mistakes

Leaving your own traffic unfiltered. Visits from your team and your agency corrupt the data. Set up the internal traffic filter and verify it is actually active — in GA4 it stays in “testing” mode by default.

Letting consent and measurement contradict each other. When a user declines, measurement must stop. A tracker that runs without consent is both a legal risk and unreliable data.

Not picking a single source of truth. The ad panel says 40 conversions, GA4 says 31, the CRM says 24. All three count different things and all three are internally correct. Decide upfront which one drives decisions — usually the CRM, because that is where the money is visible.

Setup checklist

• Key events defined, fewer than four
• Qualified enquiries separated from raw submissions
• Values assigned to events
• utm_* and gclid carried through to the CRM
• Internal traffic filter active (not in testing mode)
• Measurement consistent with the consent banner
• One source of truth chosen for decisions