Attribution Model

Every revenue figure for organic search rests on one. Google Analytics 4 offers three in its attribution reports: data-driven attribution, paid and organic last click, and Google paid channels last click, and all three exclude direct visits from credit unless the whole path is direct^[1].

The choice moves the number. Under paid and organic last click, a path of display, social, paid search and organic search gives organic search all of the credit; under Google paid channels last click, the same path gives it to paid search^[1]. Data-driven attribution shares the credit by how much each interaction changed the likelihood of the key event^[1].

Changing it rewrites the past. In GA4, a change to the reporting attribution model applies to historical and future data, and the first click, linear, time decay and position-based models have not been available since November 2023^[2]. Google Ads moved conversion actions still on those models to data-driven attribution^[3]. Which model to agree with the business, and when, is set out in the guide to SEO reporting in revenue.

A share depends on its base as a credit depends on its rule. In QueryRev’s study, the same 580 sites that block GPTBot by name are 9.2% of the 6,294 sites read, 5.8% of the 9,973 domains of the panel and 43.3% of the 1,339 sites with rules for AI crawlers^[4]. None of the three is wrong; each answers a different question, and so does each attribution model.

The model is a setting of the analytics property. In GA4 the attribution settings hold the reporting attribution model, the channels that can receive credit and the key event lookback window, which is 90 days by default for most key events, with 30 or 60 days as options^[2].

Not every report follows the model. In GA4, a change applies to key event reports that use event-scoped traffic dimensions, such as source, medium and default channel group, while session- and user-scoped dimensions are unaffected^[2].

To see what the model does to organic search, compare the channel’s key events and revenue under two models over the same period, and record the model and the window beside every figure that leaves the property.

With data-driven attribution, credit can arrive in fractions: one key event shared between two keywords shows as two parts that add up to one^[2].

Sample data: an invented site, to show the method.

A sample garden retailer records 400 orders worth $48,000 in a month. Under paid and organic last click, organic search is credited with 150 orders and $18,000. Under data-driven attribution, with the same orders and the same 90-day window, it is credited with 171.4 orders and $20,600, because its early visits now earn part of the credit. The $48,000 in the order record is known revenue, and it does not change with the model.

Neither figure is the true one. The report names the model beside the figure, keeps the same model from month to month, and labels the revenue GA4-reported until the order record confirms it.