Last updated on August 13, 2024 at 11:56 AM
Primary attribution model
There are six common attribution models: data-driven attribution, last-click attribution, first-click attribution, linear attribution, time decay attribution, and location-based attribution. We can briefly take a look at the data differences and attribution logic of different attribution models.
1. Data-Driven Attribution
Data-driven attribution uses machine learning to analyze all available data and dynamically assign conversion values to all touchpoints of engagement. This model takes into account the sequence of touchpoints, the nature of the interaction, and other user behavior data to determine each touchpoint’s contribution to conversions.
Example: Let’s say a user completes a purchase via the following path:
- First click: Google Ads -> Visited website but did not purchase
- Second click: Facebook ad -> Visited website but did not purchase
- Third click: email marketing -> visit the website and complete the purchase
In a data-driven attribution model, a conversion value might be assigned as:
- Google Ads: 20% (because it leads users to visit for the first time)
- Facebook ads: 30% (increases user interest)
- Email marketing: 50% (ultimately leads to purchase)
2. Last Click Attribution
In a last-click attribution model, all conversion values are assigned to the channel where the user last clicked before converting.
Example: Same user path:
- Google Ads -> Facebook Ads -> Email Marketing -> Complete Purchase
In the last-click attribution model, email marketing will receive 100% of the conversion value because it is the last touch point.
3. First Click Attribution
In a first-click attribution model, all conversion values are assigned to the user’s first touchpoint along the entire conversion path.
Example: Same user path:
- Google Ads -> Facebook Ads -> Email Marketing -> Complete Purchase
In a first-click attribution model, Google Ads will receive 100% of the conversion value because it is the first touchpoint.
4. Linear Attribution
The linear attribution model assigns a conversion value equally to every touchpoint along the conversion path.
Example: The user path is as follows:
- Google Ads -> Facebook Ads -> Email Marketing -> Complete Purchase
In a linear attribution model, each channel receives 1/3 of the conversion value (33.3%) because they are equally important in the conversion path.
5. Time Decay Attribution
The time decay attribution model assigns a conversion value based on the distance of the touch point from the conversion time. The closer the touch point is to the conversion time, the higher the conversion value is obtained.
Example: The user path is as follows:
- Clicked on Google Ads 5 days ago -> Clicked on Facebook ads 3 days ago -> Clicked on email marketing today -> Completed purchase
In a time decay attribution model, the possible assignments are:
- Google Ads：20%
- Facebook ads: 30%
- Email marketing: 50%
Because email marketing is closest to the actual purchase, it gets the most conversion value.
6. Position-Based Attribution
The location-based attribution model assigns 40% of the conversion value to the first and last touchpoints, and the remaining 20% evenly to intermediate touchpoints.
Example: The user path is as follows:
- Google Ads -> Facebook Ads -> Email Marketing -> Complete Purchase
In a location-based attribution model, the possible assignments are:
- Google Ads: 40%
- Facebook ads: 20%
- Email marketing: 40%
The above are our six common attribution models. Different attribution models will cause us to see different credit distributions in the conversion data. Understanding the data differences and attribution logic of different attribution models will help us understand the subsequent content.
GA4’s attribution model
GA4 currently uses data-driven attribution and last-click attribution. These two attribution models are no longer used by GA4. Therefore, the subsequent content is mainly focused on data-driven attribution and last-click attribution.
In the attribution settings of GA4, we can see that there are two attribution models to choose from: Data-driven and Last click. In the previous introduction, we saw the difference between the two attribution models, Data-driven and Last click. Data-driven is mainly driven by machine learning and data to dynamically allocate conversion credit. Last click allocates all conversion credit. All attributed to the last clicked channel.
The difference between last-click attribution in paid and natural channels and Google paid channels
Last-click attribution for paid and organic channels:
Ignore direct traffic and attribute 100% of the key event value to the channel where the customer last clicked before converting (or engaged view on YouTube).
Example: Display Advertising > Social > Paid Search > Natural Search → 100% attributed to Natural Search
Last click attribution for Google paid channels:
Attribute 100% of the key event value to the Google Ads funnel where the customer last clicked before converting.
If there are no Google Ads clicks in the path, the attribution model falls back to “last click from paid and organic channels.”
Example 1: Display > Social > Email > Direct → 100% attributed to email (fallback to last non-direct click)
Example 2: Display > Social > YouTube EVC (Engaged Views) > Email → 100% attributed to YouTube
Impact of attribution model changes
Changes to the attribution model will only affect data on key events and will not affect data related to user or sessions. In other words, the difference in the attribution model will only affect the data of “Attribution” related dimension in our exploration report. Dimensions related to users and sessions, such as First user campaign, Session camapign, etc., will not be affected. User and session related dimensions use the last-click attribution model, which we cannot modify.
User related dimensions:
Session related dimensions:
The data we see in the report is all related to users and sessions, so the data we see in the report uses the last-click attribution model.
Google Ads credit allocation settings
We can also set the credit distribution of Google ads in GA4, where one is allocated only to Google paid channels, and the other is allocated to paid channels and natural channels. If the setting is to allocate the conversions brought by Google ads only to Google paid channels, then the contributions from other channels will not be considered; if the settings are set to allocate the conversions brought by Google ads to paid channels and natural channels, then GA4 will consider the contributions of all channels. credit.
Attribution window for GA4
In addition to setting different attribution models, we can also set different attribution windows. Key events may occur days or weeks after a user interacts with your ad. The key event lookback window determines how long ago interactions were eligible for attribution. For example, a 30-day critical event lookback window means that a critical event that occurred on January 30th will only be attributed to interactions between January 1st and 30th.
Changing the key event lookback window only applies to future data. These changes will be reflected in all reports for this analytics property.
The key event of user acquisition (first_visit) can be set to 7 days or 30 days, interactive viewing is 3 days, and other key events marked as key events can be set to 30 days, 60 days or 90 days. One thing we need to note is that the new users we see in the report refer to the attribution window for key user acquisition events set here, which is 7 days or 30 days. So, a new user is a user who has not visited the website in 7 days or 30 days (depending on our settings).