What are key events in GA4?

Key events are the GA4 events you mark as most important to your business, usually purchase on a store. They used to be called conversions. GA4 counts them in its reports and shares out credit for them across your channels. If you mark the wrong ones, every channel report measures the wrong finish line.

Google's definition is short. A key event "measures an action that's particularly important to the success of your business", and any event you collect can become one. You find the event in GA4, click its star, and from then on GA4 treats it as a result.

The word conversion did not disappear. It moved. In Google's current wording, a conversion is an action you use in Google Ads to measure campaigns and bid. It is often built from a key event. The chain runs event, then key event, then conversion.

What one store's data shows

One store's anonymised GA4 export, 1 January 2024 to 21 August 2026. It holds shares of revenue only: no ad spend, no order counts.

Every row in that table hangs off a key event. On a store, revenue arrives with a purchase, so these shares describe paths that ended in one. Pick a different key event and the same visits would tell a different story.

Start with the empty rows. On the Channels sheet, Paid Social and AI Assistant hold 0.0% of revenue in all three views. Paths that touched them carried no purchase revenue, to one decimal place. If those channels sent visitors, they may have browsed, signed up or added to cart. A revenue export only credits the key event at the end.

Then the touched view. Added up, its column comes to 110.4% on the Channels sheet, by design: a journey that touched two channels counts in both. One purchase, two channels, two claims. Last click and first click give each key event to a single channel, while the touched view gives it to every channel on the path.

On the Channels sheet, Direct carries 57.7% of revenue whether you read last click, first click or touched. So more than half of the key event revenue sat on paths GA4 could only label Direct.

And most of it came fast. Journeys with 1 touch hold 79.5% of revenue and took 0.5 days to buy (Journeys sheet). Star add_to_cart and begin_checkout as key events too, and each of those quick buyers would count three times in the Key events column. They paid once.

What the export cannot show: which events this store marked, how it counted them, or what any channel caused. It is one store, not a benchmark.

Why does "conversions, renamed" mislead?

Because the rename split one word into two jobs. Key events live in GA4 and count actions from every channel, paid or not. Conversions live in Google Ads and drive bidding there, and Google says Google Ads conversions don't appear in standard GA4 reports. So one purchase can be a key event in GA4 and a conversion in Google Ads, counted under different rules.

Take the windows. GA4 looks back 90 days for a purchase key event by default. Google Ads counts a purchase within 30 days of a click unless you change it. When the two disagree about the same sale, nothing is broken: they answer different questions.

A star is also not a time machine. Google says marking an event as a key event affects reports from that moment and does not change historic data. A new counting method applies to future key events only. The reporting attribution model is the odd one out: change it, and GA4 recalculates history too.

And the defaults are generous. GA4 marks purchase as a key event by default, and a Google Ads link with ads personalization on adds more, such as add_to_cart and session_start. What that does to your attribution paths has its own answer. Here it simply means your Key events total may be counting visits.

What can a key event not tell you?

Why the action happened. A purchase key event records that someone bought, and the credit GA4 hands out depends on the attribution model you picked. Neither the count nor the credit says what would have happened without a given channel. For that you need a holdout test.

How old a touch can be and still count. A touch older than the lookback window earns nothing, whatever it did. For first_visit and first_open, the window is shorter than for purchase by default.

What it was worth. Purchase carries its own value. Any other key event needs a value and a currency. Google says an event without a valid currency is counted but not sent to Google Ads.

What to do this week

1. List your key events. In GA4, go to Admin, then under Data display click Events, and open the Key events tab. Pass: purchase, plus only actions you would pay for. Fail: session_start or view_item carries a star, so your Key events column counts browsing.
2. Watch a purchase arrive. Place a test order, then open Reports, then Realtime, and find the Key events by event name card. Pass: purchase shows within a few minutes. Fail: it never appears, so fix the tag before you trust any channel report.
3. Check the counting method. On the Key events tab, click the far right of the purchase row, then Change counting method. Pass: Once per event. Fail: Once per session, which GA4 set by default for key events created from Universal Analytics goals, so switch it and note the date.

Check the homework. Your GA4 Attribution paths export already holds the evidence. Causality Engine reads that one file and shows what each channel caused next to what last-click gave it, in 1 to 2 minutes, for €99 once (excluding VAT), refundable within 30 days. Check the homework

Sources, 1 October 2026: About key events (Analytics Help); Conversions vs. key events in Google Analytics (Analytics Help); Select attribution settings (Analytics Help); About conversion windows (Google Ads Help); Mark events as key events (Analytics Help); Change the counting method of key events (Analytics Help); Create or modify key events (Analytics Help); Realtime report (Analytics Help)

Related answers

• All GA4 how-to questions
• How to set up key events in GA4, step by step
• What key events should a home goods store track in GA4?
• What does the conversion paths report in GA4 tell me?
• Conversion, defined