GA4 Key Events and Conversions: The March 2024 Rename

On March 21, 2024, Google renamed GA4 conversions to key events. The change was platform-wide and permanent: every admin toggle, every report column, every API field that read "conversion" in GA4 now reads "key event." There is no setting to revert it. Understanding what drove this change — and what it means for how Google Analytics 4 measures your most important actions — is the foundation of accurate measurement in 2026.

The rename enforced a conceptual boundary that had always existed but was poorly communicated in Universal Analytics and early GA4. Before March 21, 2024, the word "conversion" appeared in both GA4 and Google Ads, labeling related but distinct things. Practitioners compared the two "conversions" columns and found they never matched — then spent hours debugging a discrepancy that was architectural, not a bug.

The new vocabulary resolves the ambiguity cleanly:

• GA4 side: an important on-site action measured across all traffic, all channels, all reports = key event
• Google Ads side: a key event imported into Google Ads, used to evaluate campaign performance and power Smart Bidding = conversion

You mark a key event in GA4. You then create a Google Ads conversion based on that key event inside Google Ads. The two are related but distinct measurements, attributed by separate models, scoped to different data sets, and reported in different platforms. The columns will never match exactly — and this is by design, not a misconfiguration.

This pillar covers the complete measurement layer: how the GA4 event model produces raw event data, how you promote events to key events, how GA4 attributes those key events across channels using its remaining attribution models, and how key events travel into Google Ads where they are re-attributed and counted as campaign conversions. The conversion tracking architecture in Google Ads is the downstream recipient of everything configured in GA4.

GA4 for PPC and lead generation depends entirely on key events being correctly configured, counted, and attributed before a single key event reaches Smart Bidding. A misconfigured counting method or wrong lookback window silently distorts every downstream report and every bid signal. The sections below explain each configuration layer and the non-obvious behaviors that produce measurement gaps.

Key Takeaways

• GA4 renamed "conversions" to key events on March 21, 2024. The rename is complete, permanent, and platform-wide. The word "conversion" in GA4 now refers only to a key event imported into Google Ads.
• GA4 has three attribution models in 2026: data-driven attribution (the property default), paid-and-organic last click, and Google paid channels last click. First-click, linear, time-decay, and position-based models were removed by November 2023.
• The default lookback window is 90 days for most key events and 30 days for acquisition events (first_open, first_visit). YouTube engaged-view attribution has a fixed 3-day window that cannot be changed.
• Lookback window changes are not retroactive. Historical data does not reprocess after a window change. The full benefit of a window extension takes time to appear in reports.
• The default counting method is once per event for key events created natively in GA4, and once per session for key events migrated from UA goals. These defaults apply per key event, not per property — an inherited property requires a per-event audit.
• GA4 key events imported into Google Ads via the GA4 interface default to secondary conversion action status. To use them for Smart Bidding, manually promote them to primary in Google Ads.
• GA4 attribution and Google Ads attribution are separate models trained on different data scopes. The two platforms will never report identical conversion totals. This is architecture, not a bug.

Conversions vs. Key Events: The Complete Terminology Map

The March 2024 rename introduced two parallel vocabularies that must be kept distinct. GA4 uses "key event" for what it measures. Google Ads uses "conversion" for what it optimizes. The table below maps every parallel concept so reports from both platforms can be read without terminology confusion.

The GA4 glossary hub covers the full platform vocabulary. For the purposes of this pillar, the key operational rule is: configure key events in GA4 Admin, then import them into Google Ads if campaign attribution is needed. Never assume GA4 and Google Ads will report the same count for the same underlying action.

GA4 Attribution Models Available in 2026

GA4 launched with seven cross-channel attribution models. By November 2023, Google removed four of them — first-click, linear, time-decay, and position-based — citing adoption below 3% of conversions. Three models remain. Data-driven attribution is the property-level default. Last click and Google paid channels last click are available as manual alternatives.

Important scope caveat: The attribution model set in GA4 Admin → Attribution Settings governs the Advertising section and cross-channel attribution reports. Standard acquisition reports — Traffic Acquisition and User Acquisition — follow paid-and-organic last click attribution regardless of the property-level model setting. This architectural split is a persistent source of confusion when DDA is the property model but acquisition reports appear to contradict it.

Properties using one of the four removed models before November 2023 were automatically migrated to data-driven attribution. No manual reconfiguration was required, but any reports comparing DDA against the legacy models no longer exist in the GA4 interface. The attribution models report in GA4 Admin now compares DDA against paid-and-organic last click only. MB Adv Agency treats this comparison as the standard baseline check for any property audit — the delta between DDA and last click surfaces how much upper-funnel credit DDA is redistributing away from the final touchpoint.

GA4 Attribution Lookback Windows

The lookback window determines how far back in time GA4 searches for touchpoints to attribute credit when a key event fires. GA4 sets the window at the property level, and the setting applies simultaneously to all key events and all attribution models. The defaults differ by key event type.

Retroactivity: Lookback window changes apply going forward only. Historical data does not reprocess after a window change. After extending from 30 to 90 days, reports will not show the benefit of the longer window immediately — the new window needs time to accumulate touchpoints that fall within the extended range. Plan window changes ahead of measurement reviews, not during them.

The lookback window interacts with the attribution model: DDA evaluates all touchpoints within the window to compute fractional credit. Paid-and-organic last click ignores all but the final touchpoint within the window. A 90-day DDA window captures far more multi-touch signal than a 30-day DDA window for the same property. For ecommerce tracking where purchase cycles are long, a shorter window actively understates the contribution of early touchpoints.

GA4 Conversion Counting Methods

GA4 offers two counting methods for key events: once per event (count every occurrence) and once per session (maximum one per session). The default depends on how the key event was created, not a single property-level setting. Practitioners auditing inherited properties must verify the counting method per key event — not per property.

The practical consequence of a once-per-session default on a lead generation property: if a user submits three forms in one session, GA4 records one key event, not three. This systematically understates lead volume in multi-step funnels. The inverse problem — using once-per-event on a scroll-depth key event — inflates key event counts with no business signal. Neither scenario triggers a GA4 warning in the UI.

MB Adv Agency audits counting methods as the first check on any inherited GA4 property. A property migrated from UA via the Setup Assistant goals import tool carries the once-per-session default for those migrated events while natively created events default to once-per-event. Both counting methods coexist in the same property with no visual distinction in reports. The audit requires checking the three-dot menu next to each key event in Admin → Data display → Events.

How to Mark an Event as a Key Event in GA4

Any event that GA4 has collected can be marked as a key event directly in the Admin interface. No code change is required. The toggle applies retroactively to data already in the property — the key event count in reports will reflect historical occurrences of the event, not only future ones.

Implementation context: GA4 key events fire based on the event names your Google Tag Manager setup sends. If the event has not been collected yet — because the tag has not fired — the event row will not appear in the Existing events table and cannot be toggled. The event must appear in the table before it can be marked. For new implementations, trigger the event manually in a test environment to populate the row before configuring it as a key event.

For enhanced conversions that supplement key event data with hashed first-party data, the key event must already be configured before enhanced conversions can be layered on top. Enhanced conversions add measurement accuracy; they do not replace the key event configuration.

GA4 Key Events vs. Google Ads Conversions: The Import Pipeline

When a GA4 key event is imported into Google Ads, it becomes a conversion action in Ads — but it is re-attributed by the Google Ads attribution model, not the GA4 model. The two measurements track the same underlying user actions with different credit assignments. This is the architectural reason GA4 and Google Ads conversion columns will never match exactly.

The secondary-status default: Conversions created via the GA4 interface are set as secondary in Google Ads automatically. Secondary conversions appear in reporting but do not influence Smart Bidding. This is designed to prevent double-counting when both a GA4-imported conversion and a natively created Google Ads conversion action track the same event. Practitioners who configure a key event in GA4, import it, and then wonder why Target CPA is ignoring it have hit this default. The fix is to open the conversion action in Google Ads and change the optimization setting from "secondary" to "primary." See Google Ads metrics and KPIs for how conversion columns reflect primary vs. secondary status.

For the full attribution architecture comparison between the two platforms, see the Google Ads conversion tracking and attribution pillar. The short version: GA4 DDA and Google Ads DDA are different models trained on different data scopes. Expecting them to produce the same output is a category error.

Attribution Model Selection Guide

GA4's three remaining attribution models serve different measurement needs. Selecting the wrong model produces channel-level reports that misrepresent the contribution of upper-funnel touchpoints. The table below maps scenarios to the correct model and rationale.

DDA in GA4 evaluates up to 50 actions over the lookback window to compute fractional credit assignments. When a GA4 property has insufficient volume for the model to train reliably — Google does not publish the exact threshold in official documentation — the model's outputs become less stable. GA4 does not display a warning or ineligibility flag in the UI when this occurs. The only indication is instability in the DDA-vs-last-click comparison report: large swings in channel-level credit across reporting periods suggest insufficient training data.

MB Adv Agency uses the attribution models comparison report as a standard diagnostic in quarterly measurement reviews. A DDA model that consistently credits organic search 30–40% of conversions that last-click assigns entirely to paid search indicates organic is functioning as a significant upper-funnel channel — and that paid search bids should not be evaluated on last-click ROAS alone. This is the practical reason the DDA default matters: it changes which channels appear to "earn" spend.

Enhanced Conversions in GA4

Enhanced conversions supplement GA4 key event data by matching hashed first-party customer data — email address, phone number, home address — with Google's consented, signed-in user data. The purpose is to recover key events that standard cookie-based measurement misses due to cookie blocking, Safari's Intelligent Tracking Prevention, or consent mode restrictions.

The mechanism works at the Google infrastructure level: customer data is hashed locally before transmission. Google matches the hashed value against its own signed-in user data to attribute the conversion to a known ad interaction that was not captured by the standard tag. The result appears in reports as an additional attributed conversion — the underlying key event configuration is unchanged.

For properties operating under consent mode — particularly relevant in the EU where consent rates below 70% are common — enhanced conversions recover signal that consent-blocked users would otherwise remove from reports entirely. The improvement is to measurement coverage, not to conversion rate. Enhanced conversions report what was happening before; they do not generate additional conversions. See GA4 data collection and privacy for the full consent mode interaction and the integrations and BigQuery pillar for exporting key event data for attribution modeling outside GA4's native reports.

For properties using Core Web Vitals optimization to improve landing page performance — a common parallel initiative in measurement-mature accounts — enhanced conversions provide cleaner data on whether page speed improvements actually translated to more measured key events or whether measurement gaps were masking real conversions all along.

How GA4 Data-Driven Attribution Works

GA4's data-driven attribution model uses machine learning to analyze converting and non-converting customer paths simultaneously and assigns fractional credit to each touchpoint based on its incremental contribution to the conversion. The model evaluates up to 50 actions over the 90-day lookback window for each key event. Credit is not split equally — touchpoints with higher incremental lift receive proportionally more credit.

DDA differs from rule-based models in a fundamental way: the credit assignments are not predetermined by a formula. They emerge from the data. A touchpoint that consistently appears in converting paths but rarely in non-converting paths receives high credit. A touchpoint that appears with equal frequency in both receives low credit. This means DDA outputs are property-specific — the same channel receives different credit percentages on different properties depending on how users in that property actually behave.

When DDA has insufficient data volume to train reliably, it silently shifts toward last-click-like behavior. GA4 does not display an ineligibility warning. The only diagnostic is instability in the attribution models comparison report: DDA credit that swings erratically across reporting periods indicates the model is operating on thin data. For these properties, paid-and-organic last click is the transparent choice.

Attribution Model Removal Timeline

The consolidation from seven to three attribution models in GA4 happened across a defined sequence of milestones in 2023, ending with the March 2024 key event rename that completed the platform-level terminology update.

April 2023: Google announced intent to retire first-click, linear, time-decay, and position-based models from GA4 and Google Ads, citing adoption below 3% of conversions. (Google Ads Developer Blog, April 2023)

May 2023: New conversion actions in GA4 could no longer be configured to use the four deprecated models. Existing conversion actions using them continued to function. (Search Engine Land, October 2023)

October–November 2023: Full removal completed. All four models were removed from GA4 and Google Ads. Existing properties using the deprecated models were automatically migrated to DDA. The official Google support page states "as of November 2023"; Search Engine Land coverage published October 17, 2023 reported "mid-October." The full cutover completed sometime in this window. (Search Engine Land; Google Analytics Help)

March 21, 2024: GA4 renamed "conversions" to "key events" platform-wide, completing the terminology separation between GA4-measured actions and Google Ads-counted conversions.

2026: Three models in operation — DDA (default), paid-and-organic last click, Google paid channels last click. Historical reports using the removed models no longer exist in the GA4 interface. The Reports and Explorations section of GA4 does not surface a model-comparison option for the four removed models; only DDA vs. last-click comparison is available. For properties that need advanced attribution analysis beyond GA4's native models, raw event export to BigQuery enables custom attribution modeling outside the platform.

Three Misconceptions That Corrupt GA4 Measurement

Outdated documentation and pre-2024 tutorials leave three persistent misconceptions in circulation. Each one produces incorrect configuration decisions or misread reports. The correct position for each is stated below.

Misconception 1: "GA4 still calls them conversions — the rename was optional or partial."

The rename to key events is complete and permanent in GA4 as of March 21, 2024. There is no toggle to revert it. The word "conversion" in GA4 now refers exclusively to a key event imported into Google Ads. The Admin interface, all standard reports, Explorations, and the Admin API all use "key events." Any documentation predating March 2024 that uses "GA4 conversions" is describing what is now called a key event. The setup process is unchanged; only the label differs. (Google Analytics Help, March 2024)

Misconception 2: "GA4 uses last-click attribution by default."

GA4's default attribution model for key events is data-driven attribution. Last-click is available as a manual selection but is not the default. This has been the case since GA4's general availability — the platform was designed DDA-first. The confusion persists because Universal Analytics defaulted to last non-direct click, and practitioners who migrated assume parity. Checking the current model in Admin → Attribution Settings takes 30 seconds and eliminates the assumption. (Google — Get started with attribution)

Misconception 3: "The four removed models were just deprecated — I can still access historical data under them."

First-click, linear, time-decay, and position-based models were removed from GA4 entirely by November 2023. Existing properties using those models were automatically migrated to DDA. Historical reports using the deprecated models are no longer accessible in the GA4 interface — the models do not exist as a comparison option. The only remaining comparison is DDA vs. last-click. Any practitioner attempting to compare against a "position-based" baseline in 2026 must rebuild that analysis outside GA4, using raw data exported to BigQuery and a custom model. (Search Engine Land)

MB Adv Agency encounters all three misconceptions regularly in new client onboarding. The most consequential is Misconception 2 — practitioners comparing GA4 and Google Ads conversion columns and assuming GA4 "must be wrong" because it doesn't match the last-click Ads column are comparing DDA attribution against last-click attribution and treating the difference as an error rather than a model difference. Aligning on attribution model expectations before comparing platforms eliminates the majority of "why don't these numbers match" conversations. See also: GSC Performance Report for organic traffic data that can be used to cross-reference GA4 organic channel attribution.

Frequently Asked Questions

Methodology

This pillar draws on Google's official documentation (Google Analytics Help, Google Ads Help, Google Ads Developer Blog), third-party editorial coverage (Search Engine Land, Analytics Mania, Cardinal Path, nestscale.com, Analytics Playbook / kpplaybook.com), and Ahrefs keyword data supplied by the operator for June 2026. All numeric claims trace to a cited external source. No GA4 property-level DDA data thresholds are stated as official figures — Google does not publish an explicit numeric threshold in current documentation; the official threshold for GA4-native DDA differs from the legacy Universal Analytics MCF threshold (400 conversions/month) and the Google Ads DDA threshold (200–300 conversions per conversion action), neither of which applies to GA4 property-level DDA. Attribution model removal dates reflect the October–November 2023 window confirmed by Google's support page ("November 2023") and Search Engine Land coverage ("mid-October 2023"); the broader window is the accurate characterization. Reviewed by MB Adv Agency, June 2026.