Attribution in GA4 assigns credit for key events (GA4’s current name for what reports used to call conversions) across the traffic that drove them. As of August 2026 it uses exactly three models: data-driven attribution, paid and organic last click, and Google paid channels last click. The old rules-based models (first click, linear, time decay, position-based) were retired in November 2023, so any guide still describing them is out of date.
This is the working guide: the models themselves, where the settings actually live, the lookback windows that quietly shape your numbers, and GA4’s honest limits. Facts verified August 2026 against Google’s own documentation.
The three GA4 attribution models (and the retired ones)
The current menu has three entries, and only three.
Data-driven attribution, labeled “Data-driven (recommended)” in the chooser, is the interesting one. Its algorithms compare converting and non-converting paths to work out which touches actually mattered, weighing factors like time to conversion, device type, and the order of ad interactions. It is not a fixed rule. It is a model estimating contribution from your own data.
Paid and organic last click gives 100 percent of the credit to the last channel the customer clicked through before converting. Direct traffic is ignored, with one exception: if the entire path is direct visits, direct gets the credit.
Google paid channels last click gives 100 percent of the credit to the last Google Ads click in the path. If there is no Google Ads click at all, it falls back to paid and organic last click.
That is the whole list. First click, linear, time decay and position-based no longer exist in GA4. If a report template, agency deck or blog post still references them, it was written before November 2023 and should be treated accordingly.
Key events, conversions, and the vocabulary shift
GA4 renamed its core metric, and the confusion has never fully settled.
A key event measures an action that is important to your business: a purchase, a signup, a demo request. Key events are what appear in standard GA4 reports. They are what everyone used to call conversions.
A conversion, in current GA4 vocabulary, means something narrower: a Google Ads conversion created from a key event, used for ad measurement and bidding. Conversions live only in the Advertising section, not in standard reports.
The flow runs one way: an event becomes a key event when you mark it as one, and a key event can then be shared with Google Ads as a conversion. So every conversion starts life as a key event, but not every key event becomes a conversion.
The practical consequence is reconciliation pain. The number in your standard reports (key events) and the number Google Ads shows you (conversions) are computed in different places under different rules, with different attribution and different windows. They will not match, and that mismatch is by design, not by error. Knowing which vocabulary a stakeholder is using is half the job of explaining a GA4 report.
Where the settings live, and the windows that shape numbers
Attribution settings live under Admin, then Data display, then Events, where you will find the Key events attribution screen. Changing the reporting attribution model requires Marketer-level permissions or above, and the change applies to all key event reports and to explorations using event-scoped traffic dimensions: source, medium, campaign, default channel group.
Lookback windows are the quiet rule-makers. Acquisition key events (first open and first visit) default to a 30-day window, with a 7-day option. All other key events default to 90 days, with 30-day and 60-day options. A conversion path longer than your window simply loses its early touches.
There is also a channels-that-can-receive-credit setting with two options: Google paid channels, or paid and organic channels. Paid and organic is the default.
Here is the important bit: none of these settings are neutral knobs. A lookback window is a definition of what counts as influence. Change it from 90 days to 30 and you have not tuned a report, you have redefined the question, and GA4 will re-answer history under the new rules. Set the windows deliberately, write them down, and never compare numbers across a settings change without saying so.
GA4’s limits
Within what Google can see, GA4’s data-driven attribution is genuinely good. The boundary is what Google can see and of course how much engagement your brand drive outside of the platform’s control.
GA4 observes your website and app, plus the ad touches Google knows about. It cannot see views on other advertising platforms, journeys that hop devices without signing in, offline moments like TV or shelf, or activity inside walled gardens. When those invisible touches help drive a conversion, the credit does not vanish. It redistributes among the touches GA4 can see, quietly flattering them. The channels in the report look a little better than they are, precisely because the report cannot see their helpers.
It is also a black box. You cannot inspect the model’s weights or ask why a particular path was credited a particular way. You are trusting the process. Of course Google is notorious for developing the most cutting edge technologies especially in marketing measurement. After all, Google built most of the internet infrastructure the way we know it today.
None of this makes GA4 attribution useless. It makes it one good witness rather than the whole court. The full argument for how algorithmic attribution should be read, and where it earns trust, is in our guide to data-driven attribution.
Using GA4 attribution well
The playbook is short.
Pick the data-driven model consciously and leave it stable. Model-hopping makes every trend line a mystery.
Know your lookback windows before comparing periods. A quarter measured under a 30-day window is not comparable to a quarter measured under 90. This depends on your product or industry. If you operate an ecommerce website and the buying cycles are short maybe you can even look at shorter lookback window. Same applies if your industry is price sensitive.
Keep the vocabulary straight when reconciling against Google Ads: key events in standard reports, conversions in the Advertising section, computed under different rules. The gap between them is expected.
If you want help reading your GA4 numbers honestly, or building the layer above them, talk to MarSci.
Frequently Asked Questions
Yes, there are. Many 3rd party providers offer similar technologies. MarSci Analytics for instance, has developed its own data-driven attribution which provides a holistic view of your cross-channel activity. It works without UTMS and it doesn’t require any set up. Read more on Causal-driven attribution (CDA).
Exactly three as of August 2026: data-driven attribution (labeled recommended), paid and organic last click, and Google paid channels last click. The old rules-based models, first click, linear, time decay and position-based, were retired in November 2023 and no longer exist anywhere in GA4, so guides still describing them are out of date.
A key event measures an action important to your business and appears in standard GA4 reports. A conversion now specifically means a Google Ads conversion created from a key event, used for ad measurement and bidding, and shown only in the Advertising section. Every conversion starts as a key event, but not every key event becomes a conversion. A key event requires manual setup within the GA4 enviroment.
Under Admin, then Data display, then Events, where the Key events attribution screen holds the reporting model, lookback windows and credit-eligible channels. Changing the reporting model requires Marketer-level permissions or above, and the change applies across key event reports and explorations using event-scoped traffic dimensions like source, medium and campaign.
Acquisition key events (first open and first visit) default to a 30-day lookback with a 7-day option. All other key events default to 90 days, with 30-day and 60-day options. Windows are rules, not facts: they define which touches can receive credit, so know yours before comparing periods or accounts.
Within Google’s view it is genuinely good, but it cannot see other platforms’ ad views, offline journeys or walled gardens, so credit quietly concentrates on the touches it does see. For budget-grade decisions across channels, pair it with aggregate methods like marketing mix modeling and causal attribution that read the whole mix.
Google’s GA4 attribution aims to give the right credit to the right touchpoint considering the amount of information the model has. If channels don’t generate clicks or web / app sessions then GA4 can’t give the credit to the right touchpoint.
