Attribution models: how GA4 and Google Ads assign conversions and which model to choose

As of: 12 November 2026. We describe attribution models according to the Google Analytics and Google Ads help pages from this date. Google changes the list of available models, so compare the names with those in your account before you change any settings.

What is an attribution model?

An attribution model is a rule, a set of rules or an algorithm that decides how credit for a conversion is split between the ads, clicks and other touchpoints on the customer's path. That is how the Google Analytics help article “Get started with attribution” defines it. Attribution answers the question of which campaign brought in a sale or a lead when the customer saw several ads before deciding.

Google Ads Help illustrates it with a restaurant. A customer clicks ads after four consecutive searches, from a general query about restaurants in Tuscany to a query with the restaurant's name, and books a table after the last click. In the last click model the last keyword gets the whole conversion. In the data-driven model all four keywords share the credit, each according to its contribution.

The attribution model you choose decides which campaigns look profitable in reports, and therefore where you move budget. The same month viewed in two models can lead to opposite conclusions about the same campaign.

Which attribution models are available in GA4 and Google Ads?

GA4 offers three attribution models: data-driven, paid and organic last click, and Google paid channels last click. Google Ads offers two models: data-driven, which is the default for most conversion actions, and last click.

The rule-based first click, linear, time decay and position-based models have been retired. Google Analytics Help states that they have not been available in GA4 since November 2023, and Google Ads Help states that conversion actions using them were upgraded to data-driven attribution. Guides that describe them refer to settings you will no longer find in your account.

All models in the GA4 attribution reports exclude direct visits. The exception is a path made up only of direct visits: then direct traffic gets the whole conversion.

How does data-driven attribution work?

Data-driven attribution compares the paths of customers who converted with the paths of people who did not, and calculates how much each ad interaction changes the probability of a conversion. According to Google Analytics Help, the model considers the time from interaction to conversion, the device type and the order of ad exposures.

Google Analytics Help illustrates it with a path of four ad exposures. With all four exposures the probability of a conversion is 3%, and without the fourth it drops to 2%. The model therefore concludes that the fourth exposure raises the conversion probability by 50%, repeats the calculation for every exposure on the path and turns the results into the weights it uses to split credit.

That is why fractions appear in conversion columns. When a customer clicks ads for two different keywords and buys, each keyword gets part of the conversion, and both parts add up to 1.0.

Google Ads Help on data-driven attribution states that you can turn the model on for any conversion action regardless of its volume. Google still recommends at least 200 conversions and 2,000 ad interactions in supported networks within 30 days, because with less data the model is worse at recognising patterns. The same page notes that with limited data the data-driven and last click models can give the same result.

When does the last click model still make sense?

The last click model works when the customer path is short or when you compare results with historical data counted the same way. It gives the whole conversion to the last-clicked ad, so it is easy to check, yet it ignores the touchpoints that built interest earlier.

In our practice we choose the last click model in three situations:

• Short purchase path: the customer decides after one search, for example with an urgent local service where they call from the first ad.
• Single-click paths: when most customers click only one ad, both models show similar numbers, and last click is easier to explain.
• Comparison with old reports: when you compare results with a period in which the account counted conversions by last click.

The weakness of the last click model shows in brand campaigns. When a customer first clicks an ad for a generic query and a few days later types the store name and clicks a brand ad, the model gives the whole sale to the brand campaign. A similar mechanism inflates Performance Max results, which we describe in the article on ROAS in Performance Max.

Why do GA4 and Google Ads show a different number of conversions?

GA4 and Google Ads show a different number of conversions from the same campaign because they count them with different attribution settings. Three differences come from the attribution settings themselves:

• Channels that can receive credit: with the paid and organic channels setting, channels such as organic search and email can also get credit, while with the Google paid channels setting only Google Ads ads can. According to Google Analytics Help, Google Ads reports use Google paid channels by default, while GA4 reports use paid and organic channels.
• Dimensions in GA4 reports: the model applies to event-scoped dimensions such as source and campaign. Session-scoped and user-scoped dimensions, such as session source, do not depend on the chosen model, so the traffic acquisition report can show a different split than Google Ads.
• Scope of the setting: in Google Ads you set the model separately for each conversion action, while in GA4 one reporting attribution model covers the whole property.

Timing adds to this. In GA4 the default lookback window is 90 days for most key events and 30 days for acquisition events (first_visit and first_open), and touchpoints before the window get no credit. Google Analytics Help also states that with data-driven attribution conversions can be reattributed for up to 7 days after the conversion, so last week's result can still change. In our practice the sum of conversions reported by several ad platforms usually exceeds the number of orders in the store, because each platform counts them in its own model and credits itself.

The split in the last row is arbitrary. The data-driven model calculates weights from the data of a specific account and a specific conversion action, so the proportions in your account will differ.

How do you choose an attribution model for your account?

In our practice, for most accounts running campaigns across several Google networks the starting point is the data-driven model, because Google sets it as the default and it splits credit between earlier and later ad touchpoints. Leave last click to accounts with a short purchase path or a very small number of conversions.

Before you change the model, check the effect in the model comparison report. According to Google Ads Help, the “Model comparison” report shows two models side by side, and Google recommends testing a new model before switching and assessing its effect on return on investment.

Changing the model in Google Ads also affects bids. The model changes the “Conversions” column that target CPA and target ROAS strategies use, so after the switch the algorithm optimises for conversions counted differently. We explain how to choose the strategy itself in the article target ROAS or target CPA.

Where do you change the attribution model in GA4 and Google Ads?

In GA4 you change the attribution model in the attribution settings in Admin, and in Google Ads in the settings of each conversion action separately. The paths according to Google's help pages:

• GA4: Admin → Data display → Events → Attribution settings. The change requires the Marketer role or above at property level.
• Google Ads: Goals → Conversions → Summary, then the chosen conversion action → Edit settings → Attribution model.

In GA4 a change of the reporting attribution model applies to historical and future data, while a change of the lookback window applies from the moment you save it. According to Google Analytics Help, the model does not change session-scoped and user-scoped dimensions such as session source, so changing the model alone does not change traffic acquisition reports.

What won't an attribution model tell you about your campaigns?

An attribution model splits credit only for the conversions and touchpoints you measure, so it will not show the contribution of unmeasured channels or sales that would have happened without the ad. The question of how many extra sales a campaign brings is answered by a test with a control group.

Most often the problem lies in the definition of a conversion itself. In the Maxi Mobil case study Google Ads counted every click on the “book a visit” button as a conversion, although nobody knew how many of those clicks ended in a phone call. After call tracking and offline conversions were implemented, the cost per call fell from PLN 100.58 to PLN 51.64, a drop of 48.6%.

Before you change the model, check three things:

• Conversions: whether the conversion is a purchase or a lead rather than a button click.
• Traffic tagging: whether links from emails and social media carry UTM parameters, because without them some visits end up as direct traffic.
• Off-site sales: whether phone calls and deals from your CRM return to Google Ads as offline conversions.

You can check whether purchases and leads reach GA4 at all with the free GA4 Audit. Enhanced conversions also improve how accurately conversions are matched to clicks.

We configure attribution and conversion measurement as part of our GA4 implementation service. As a digital marketing agency we combine it with running Google Ads campaigns that rely on this data.

What sources is this article based on?

Model definitions, the description of data-driven attribution and the GA4 settings come from Google Analytics Help: “Get started with attribution” and “Select attribution settings”. The rules in Google Ads are described in the Google Ads Help articles “About attribution models” and “About data-driven attribution”, including the recommendation of 200 conversions and 2,000 interactions in 30 days. The cost per call comes from our Maxi Mobil case study.