How a key event works
Everything a visitor does in GA4 arrives as an event — a page viewed, a video started, a file downloaded, a form sent. All of them are collected the same way and none of them is special until you say so. Marking an event as a key event is that decision: it tells GA4 that this particular action is one of the things the website exists to produce.
Nothing new is tracked when you do it. The same event keeps arriving; it simply starts appearing in the key event columns and metrics that sit across GA4’s reports, so you can compare channels, campaigns and landing pages by outcome instead of by traffic.
The name is recent. Google renamed this in GA4 from “conversion” to “key event”, which is why older guides, courses and screenshots use the two words for the same thing. In Google Ads the word conversion is still used, so both terms remain in circulation and you will meet them side by side.
Why key events matter
They are the line between traffic reporting and business reporting. Sessions tell you how busy the site was. Key events tell you how often it did its job, and that is the only figure worth putting next to spend when you decide which channel deserves more budget.
They also feed decisions made by machines. When a GA4 key event is imported into Google Ads and used as a conversion action, automated bidding starts buying more of whatever produced it. A carelessly chosen key event is not just a reporting nuisance — it quietly redirects money.
Where key events go wrong
The usual error is marking too much. Page views, scrolls, outbound clicks and video plays all get flagged because they are easy to fire and they make the reports look busy. The result is a dashboard where genuine enquiries are lost among activity, and a bidding system trained to chase browsing.
The second is double counting. If the same submission fires both an automatic event and one you built yourself, every enquiry lands twice. The fix is to check in real time, submit a test yourself, and confirm exactly one event arrives.
The third is assuming GA4 and Google Ads will agree. They attribute differently, they use different time windows and they count differently, so the two totals will not match. That is expected, not a fault to be debugged.
Getting it right
Decide your outcomes before you touch the interface: usually a form sent, a call started, a message opened on WhatsApp or Viber, and a purchase. Give each one a clear, stable name, agree it with whoever reads the reports, and only then mark it. Keep the list short enough to remember.
Test each one from a real device on the site as a visitor would use it, then check it appears once. Record the date you changed anything, because a definition changed midway through a quarter will look like a performance change later. If your measurement is being built from scratch, that ordering — outcomes first, tags second — is the core of a sound GA4 setup.
