Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Marketing Analytics › Google Analytics 4

Key Events and Conversions in GA4

What GA4 key events are, how to choose and mark them, counting method and value, how they become Google Ads conversions, and why platform counts differ.
Guideintermediate10 min readUpdated

What a key event is (and the naming change)

Every event a property collects is equal until someone marks it. The mark is set in Admin under Key events, either by typing an event name — which works before the event has been seen, so a key event can be prepared ahead of a release — or by switching the toggle on an existing event in the Events list. From that point the event is counted as a key event in every report, appears in the key events report under Engagement, is offered as a column in the acquisition and traffic reports, filters the attribution and conversion path reports, and becomes eligible for import into a linked Google Ads account. Unmarking it reverses all of that going forward; history is not recalculated either way. A property may hold a documented number of key events, and the limit is deliberately small.

The product used to call these conversions. Google renamed them key events and now uses the word conversion, inside GA4, only for key events that have been shared with Google Ads and are counted there under Ads' rules. The rename recognises something practitioners had learned painfully: the GA4 count and the Google Ads count of the same outcome are produced by different systems with different attribution, windows and timing, and giving them the same name invited people to reconcile numbers that were never meant to match. The two numbers are explained below; for now, the vocabulary is key events in GA4, conversions in Ads.

Choosing key events: few, meaningful, measurable

The temptation is to mark everything that sounds valuable. The discipline is three tests. A key event should be meaningful: it should be a business outcome, or the closest reliably observable proxy for one — a purchase, a qualified enquiry, a trial start, a booking — and the measurement plan that produced the property's KPI list should already name it. A page view of a pricing page is a behaviour, not an outcome. It should be measurable: it fires once per real occurrence, from a tag or data layer push that has been tested, with the parameters a report needs, and it does not fire on refresh, on back navigation or on a test order. An event that fires inconsistently is worse as a key event than as an ordinary event, because the mark lends it authority it has not earned. And there should be few of them, because each one competes for attention in reports, each one can be imported into Google Ads where it will shape bidding, and the property's limit is finite.

A useful distinction is between the outcome itself and the steps towards it. Mark the outcome. Keep the steps — basket adds, checkout starts, form starts — as ordinary events for funnel analysis in explorations, and resist marking them so that the key event count stays a count of outcomes. Where Google Ads will import the key events, mark the primary outcome and perhaps one earlier proxy for campaigns that cannot yet accumulate enough outcomes, and set the proxy as a secondary goal in Ads so it informs reporting but not bidding.

Counting method and value

Each key event has a counting method, chosen from the key event's menu in Admin: once per event, the default, or once per session. Once per event counts every occurrence; once per session counts the first occurrence in a session and ignores the rest. Purchases and other transactions want once per event, because two orders in one visit are two outcomes. Leads, sign-ups and enquiries almost always want once per session, because a double-clicked submit button, a page refresh on the thank-you page or a person correcting a typo and resubmitting is one enquiry, not two. The setting applies to data from the moment it is changed and does not restate history, so set it at the time of marking. Reports that count sessions with a key event — the session key event rate, the conversion rate (CVR) most people mean — are unaffected by the method, since a session either contained the key event or did not; the raw key event count is what changes. The page where most purchase and lead events fire has a job beyond counting, and the guide to confirmation and thank-you pages covers what it should confirm and offer as well as how to count each conversion once.

Value comes from the event. A purchase carries value and currency from the order; a lead event can carry a value the business assigns to an enquiry. Where an event carries no value, Admin allows a default value with a currency to be set on the key event, and it is applied to every occurrence without one; values are converted into the property's currency for reporting. Value is what makes revenue and return reporting possible in GA4 and what a value-based bidding strategy optimises towards once the key event reaches Google Ads, so a default value on a lead event should be a considered estimate of what an enquiry is worth rather than a round number chosen to make the reports look full. The conversion rate calculator turns key event and session counts into the rates these decisions are discussed in.

Key events as Google Ads conversions

Once the property and the Google Ads account are linked — the Google Ads integration tutorial covers the link and its settings — any key event can be imported into Google Ads as a conversion action. In Google Ads the import is started from the conversions section by choosing to import from a Google Analytics property, selecting the web key events to bring across, and completing the action settings: primary or secondary goal, the category, the attribution model for the action and the counting and window settings Ads applies. The imported action then behaves like any other Google Ads conversion action: it can be a primary goal that bidding optimises towards, it reports in the Ads interface, and it carries the value GA4 sent or defaulted.

Three things about the import are easy to miss. The action's attribution model is set in Google Ads and is separate from the GA4 property's own attribution settings, so the same key event can be attributed differently in the two products; the attribution settings guide covers the GA4 side. Counting and windows are set on the Ads action as well, so check them after import rather than assuming they mirror GA4. And import is one of two ways to measure in Ads: the native Google Ads tag, set up in the conversion tracking tutorial, is the other, and running both as primary goals for the same outcome double-counts. The usual pattern is one primary measurement per outcome and the other as a secondary action for comparison.

Platform conversions vs GA4 key events

The GA4 key event count for a period and the Google Ads conversion count for the same imported action will not agree, and the gap is structural rather than a fault. GA4 counts every key event from every channel on the day it occurred and attributes it across paid and organic touchpoints under the property's Attribution settings. Google Ads counts only the key events it can tie to an ad click or engagement, dates them to the time of that click, applies its own click-through and engaged-view windows, can include conversions modelled where consent limited observation, and attributes them under the model set on the action. The imported action is also subject to the link's own processing, so it may lag the GA4 report. A GA4 export filtered to the paid channel, compared with Ads over a window long enough for the attribution windows to close, usually brings the two within a tolerable distance; what remains is the difference between a web analytics system and an advertising system answering different questions.

The full list of causes and a reconciliation method are in why Google Ads and GA4 disagree for this pair of systems and in the platform versus analytics discrepancies guide for the general case. The practical rule is to use each number for its own decision: the Ads number for bidding and for judging Ads performance against its own history, the GA4 number for comparing channels and for the business's view of outcomes, and neither as the arbiter of the other.

Illustrative example — not a real result

A furniture retailer's key events over a month

A furniture retailer marks two key events: purchase, counting once per event with the order value, and generate_lead for its design consultation form, counting once per session with a default value of £40. In October GA4 records 1,260 purchases and 415 consultation leads across all channels. The paid search channel accounts for 388 purchases in the GA4 acquisition report. Google Ads, which imports both key events, reports 352 purchase conversions for the same month — 36 fewer — because 29 of the GA4 purchases were attributed by Ads to clicks in late September and dated there, and the data-driven model on the Ads action gave partial credit for others to campaigns outside the comparison. The lead action shows 131 conversions in Ads against 148 sessions with the key event in GA4's paid channel. The team treats the Ads figures as the bidding signal and the GA4 figures as the channel comparison, and stops trying to make them equal.

Reporting on key events

Inside GA4, key events surface in several places and each answers a different question. The key events report under Engagement lists every marked event with its count and value over time, and is the place to watch for a key event that stops firing. The acquisition and traffic reports carry a key events column with a selector to show one key event or all of them, alongside the session key event rate and the user key event rate, which is how channel and campaign performance is compared. The Advertising section holds the attribution reports — model comparison and conversion paths — filtered by key event, which is where the question of which channels contributed, rather than which was last, gets its answer. Explorations expose a key events metric and an is-key-event dimension for free-form work, and a funnel can end on a key event. Realtime shows key events as they arrive, which makes it the first check after a release.

Two habits make this reporting trustworthy. Compare the key event count against the system of record — the order management system, the CRM's enquiry count — every month, and investigate drift beyond what consent refusals and blocked tags explain. And report rates rather than counts when comparing periods or channels, because a count moves with traffic and a rate moves with the thing marketing controls. For how key events feed a wider conversion measurement setup across platforms, the author's conversion tracking expertise page describes the working approach.

Common mistakes

  • Marking behaviours as outcomes. Scroll depth, pricing page views and video plays as key events, inflating the count and diluting the bidding signal once imported.
  • Once per event on lead forms. Double submissions and thank-you page refreshes reported as two leads each.
  • No value on anything but purchases. Lead events imported without a value, so value-based bidding has nothing to work with and return reporting stops at e-commerce.
  • Two primary measurements of one outcome in Ads. The native tag and the GA4 import both set as primary goals, doubling the conversions bidding sees.
  • Expecting the Ads and GA4 numbers to match. Days spent reconciling differences that are structural.
  • Marking before the event is stable. A key event imported into Ads while the tag still misfires trains bidding on noise and takes weeks to unlearn.
  • Never reviewing the list. Key events from a campaign two years ago still marked, still imported, still competing for attention.
  • Reporting counts, not rates. A rise in key events celebrated when traffic rose and the rate fell.

The GA4 hub explains the event model these marks sit on, and the events and parameters guide is where the events themselves are designed before any of them deserves the mark.

Frequently asked questions

Is a GA4 key event the same thing as a conversion?

Inside GA4 the thing you mark is now called a key event, and it is what the product used to call a conversion. Google reserves the word conversion for key events that have been shared with Google Ads and are counted there under Ads' own rules. The rename exists precisely because the two numbers were never the same and people kept comparing them.

Which counting method should a key event use?

Once per event when every occurrence is a separate outcome, which is the case for purchases and most transactions. Once per session when a repeated firing in one visit is the same outcome, which is the case for most lead forms, sign-ups and enquiries where a double submission or a page refresh should not count twice. Set it when you mark the event, because a later change applies going forward only.

Why does Google Ads report fewer conversions than GA4 shows key events?

Because Google Ads only counts the key events it can attribute to its own clicks or impressions, within its own windows, and dates them to the click rather than the event, while GA4 counts every occurrence from every channel on the day it happened. The imported action also has its own attribution model in Ads. Compare the GA4 number for the paid search channel against Ads, over a period long enough for the windows to close, before concluding something is broken.

Terms used on this page

Tools to try

Frameworks

Where this page is used

Also related

Sources

  1. Google Analytics Help — Google Analytics Help

Related