Attribution in Google Analytics 4: Models, Settings, and Interpretation

A typical customer visits your site three or four times before converting: an Instagram ad, then a Google search, then an email. Who gets credit for the conversion? That’s the question of theallocation. The answer determines which channels appear to be profitable—and therefore where your budget goes. GA4 offers a simplified attribution system compared to Universal Analytics; here's how it works and how to interpret it correctly.

What the Assignment Measures

For each conversion, GA4 records the sequence of sessions that preceded it, along with their sources. The attribution model determines the credit assigned to each touchpoint. Let’s take the example of a €100 purchase preceded by three visits: a Meta ad, an organic Google search, and an email.

The two available models

Data-driven

The default model. It uses machine learning to estimate the actual contribution of each touchpoint based on all the paths through your property. It compares paths that convert with those that don’t and assigns greater value to the steps that increase the probability of conversion. Advantage: It reflects your actual situation rather than an arbitrary rule. Limitation: It is not very transparent (it’s not clear exactly why a channel receives 35%) and requires a large volume of data to be reliable.

Last click

All credit goes to the last channel before conversion, excluding direct traffic (if the last touchpoint is “direct,” GA4 goes back to the previous click). It’s simple and comparable to what Google Ads does by default, but it underestimates discovery channels (social, display, content) and overestimates end-of-journey channels (brand, email, searches for your name).

Details on how the default model works can be found in our fact sheet on the data-driven attribution model.

Configure the template and windows

Administration > Assignment Settings allows you to choose:

• The Report Template : based on data or last-click (with or without a preference for Google paid channels). The change applies retroactively to all reports.
• Allocation windows : 30 days by default for acquisition conversions (first_open, first_visit), 90 days for other key events, adjustable to 7, 30, 60, or 90 days. A short window is suitable for impulse purchases, while a long window is better for B2B.
• Channels Considered : By default, all; the "Google paid and organic channels" option limits attribution to Google clicks.

Where to view attribution in GA4

1. Advertising > Attribution > Conversion Paths : actual conversion paths (for example, “Organic Social > Organic Search > Direct”), along with the number of conversions and the value for each combination. The “Initial / Intermediate / Final Touchpoints” columns show the role of each channel.
2. Advertising > Attribution > Model Comparison : the number of conversions per channel according to the data-driven model and the last-click model, side by side. This report reveals an underestimated channel: if social media goes from 40 conversions in the last-click model to 95 in the data-driven model, it plays a discovery role that the last-click model fails to capture.
3. Traffic Acquisition : Key events are assigned based on the selected reporting model; the “User Acquisition” report uses the user’s first source.

The Pitfalls of Attribution

• Compare GA4 and Google Ads. Google Ads attributes conversions to the date of the click using its own window and counts view-through conversions; GA4 attributes conversions to the date of conversion. The numbers always differ. See Connect GA4 to Google Ads.
• Direct traffic. It captures unidentifiable visits (bookmarks, apps, links without referrers). A high percentage of “direct” traffic among recent visits often means that the customer is returning after having been persuaded elsewhere.
• Offline channels. Word of mouth, trade shows, and cold calling: these are not tracked by GA4. The model allocates 100% of the credit to digital channels, which automatically overestimates their impact.
• Consent. The user journeys of users who refuse cookies are not reconstructed, or are only modeled. See GA4 and GDPR.
• Campaigns without tags. An email without UTM parameters appears as "direct" and disappears from the attribution report. That's why it's important to tag your campaigns.

What should we do with this information?

Attribution doesn’t reveal a single truth; it reveals a distribution. Use it to make three decisions: maintain a budget for a discovery channel that appears unprofitable based on last-click attribution; identify common channel combinations (social followed by SEO) and support them together; set cost-per-conversion goals consistent with each channel’s role. The specific case of paid social media—which is often penalized by last-click attribution—is discussed in GA4 and SMA: Properly Attributing Performance.