The trouble with GA4 is not that it tracks too little. It tracks a great deal, automatically, and almost none of what it gathers on its own helps decide anything.
Scrolls, outbound clicks, site searches, video: they arrive free and fill the reports. Then somebody asks what a customer cost this month and there is no way to answer, because nobody ever defined what a customer is.
The six events that matter
| Event | What it says | When it fires |
|---|---|---|
| key page view | real interest | pricing or services page seen |
| form start | intent | first field filled |
| form submit | a lead | successful submit, not a button click |
| contact | a conversation | reply to a call or email |
| purchase | revenue | order confirmed |
| lead quality | the truth | the lead was good, judged by a person |
The last one is the one nobody configures and the one worth more than all the others.
Without it you optimise campaigns toward leads, and bad leads count as much as good ones: you end up funding the channel producing the most filled forms rather than the one producing customers.
Without sending lead quality back into GA4, you are optimising toward noise.
Successful submit versus button click
The most widespread technical mistake, and the most expensive.
Many people track the click on the "Submit" button. But that click also happens when the form then fails on an invalid field, on a server error, or because somebody clicked twice. The result is a lead count higher than reality, inflating every calculation downstream.
The event has to fire after submission is confirmed: on the thank-you page, or on the form's success response. Unable to tell those apart, your cost per lead is understated and you never noticed.
Pixel, GA4 and conversion API: what does what
- AIGA4 — tells you what happened on your site. Your archive, not Meta's
- AIThe Pixel — tells Meta who converted, so it can find similar people
- AIConversion API — sends the same event from the server, so it survives browser blocking
- NOIYour CRM — the only one that knows whether that lead ever bought
All four are needed and they do different jobs. The mistake is assuming one replaces the others: the Pixel gives you no historical archive, GA4 does not optimise campaigns, and neither knows whether the contract got signed.
How to run these checks without doing them by hand is in the AI agents guide.
How to verify it, in half an hour
Before switching any campaign on, not after.
Open GA4's realtime report and fill the form for real, from a phone, in a private window. One event should appear, not two. Then check the event still arrives with an ad blocker running, which is the condition a sizeable share of visitors are in. Then confirm the monetary value attached to the event is not zero, or every return report stays empty.
And finally the part almost nobody does: wait twenty-four hours and check the event count in GA4 resembles the number of emails you actually received. When GA4 says 40 and the inbox holds 12, there is a leak to find before spending anything.
What is not worth tracking
Scroll depth, unless you are studying one specific long page. Social clicks. Time on page, which GA4 measures badly by design. And every micro-interaction that will not change a decision.
Each extra event is noise the six real ones drown in, and the deciding question never changes: if this number doubled, what would I do differently? With no answer, the event does not belong.
What to do with the numbers once they are trustworthy is in the funnel guide.
Frequently asked questions
Which events should you configure in Google Analytics 4?
Six: a key page view, form start, form submit, contact, purchase, and a lead-quality event judged by a person. GA4 collects plenty automatically, but what it gathers on its own measures traffic rather than results, so the events that matter have to be defined.
Why is tracking the submit button click not enough?
Because that click also fires when the form then fails, on an invalid field, a server error or a double click. The lead count comes out higher than reality and inflates everything downstream. The event should fire after submission is confirmed, on the thank-you page.
What is the difference between GA4, the Pixel and the conversion API?
GA4 is your own archive of what happened on the site, the Pixel tells Meta who converted so it can find similar people, and the conversion API sends the same event from the server so it survives browser blocking. All are needed, and none replaces the others.
How do you verify tracking works?
By filling the form for real, from a phone in a private window, watching the realtime report: exactly one event should appear. Then repeat with an ad blocker running. And after a day, compare the event count against the emails actually received: a mismatch means a leak.
What does tracking lead quality mean?
Sending back into GA4 the fact, judged by a person, that a given lead was good. Without it you optimise campaigns toward filled forms and bad leads weigh as much as good ones: you end up funding the channel bringing the most contacts rather than the most customers.
Which events are not worth tracking?
Scroll depth outside specific cases, social clicks, time on page which GA4 measures badly by design, and every micro-interaction. The deciding question never changes: if this number doubled, what would I do differently? With no answer, the event is just noise.
Want to know whether your tracking tells the truth?
We will check what actually reaches GA4 and where it leaks. Seven questions, founders answer within 48 hours.
ApplyKeep reading
Founder of As Growth, an AI-first growth partner. He runs growth and strategy across partner businesses: OroEtic, Giojello, Equity Plus and others. Apply if you want to work with us.