Attribution models and lookback windows
The concept of attribution in Adobe Analytics requires two components:
- Attribution model: The model describes the distribution of conversions to the hits in a group. For example, first touch or last touch.
- Attribution lookback window: The lookback window describes which groupings of hits are considered for each model. For example, visit or visitor.
Attribution models
2^(-t/halflife)
, where t
is the amount of time between a touch point and a conversion. All touch points are then normalized to 100%.Lookback windows
A lookback window is the amount of time a conversion should look back to include touch points. Attribution models that give more credit to first interactions see larger differences when viewing different lookback windows.
-
Visit lookback window: Looks back up to the beginning of a the visit where a conversion happened. Visit lookback windows are narrow, as they don’t look beyond the visit. Visit lookback windows respect the modified visit definition in virtual report suites.
-
Visitor lookback window: Looks at all visits back up to the 1st of the month of the current date range. Visitor lookback windows are wide, as they can span many visits. Visitor lookback considers all values from the beginning of the month of the report’s date range. For example, if the report date range is September 15 - September 30, the visitor lookback date range includes September 1 - September 30.
-
Custom lookback window: Allows you to expand the attribution window beyond the reporting date range, up to a maximum of 90 days. Custom lookback windows are evaluated on each conversion in the reporting period. For example, for a conversion occurring on Feb 20th, a lookback window of 10 days would evaluate all dimension touchpoints from Feb 10th to 20th in the attribution model.
Here is a video on custom lookback windows:
embed
Example
Consider the following example:
- On September 15, a visitor arrives to your site through a paid search advertisement, then leaves.
- On September 18, the visitor arrives to your site again through a social media link they got from a friend. They add several items to their cart, but do not purchase anything.
- On September 24, your marketing team sends them an email with a coupon for some of the items in their cart. They apply the coupon, but visit several other sites to see if any other coupons are available. They find another through a display ad, then ultimately make a purchase for $50.
Depending on your lookback window and attribution model, channels receive different credit. The following are some notable examples:
-
Using first touch and a visit lookback window, attribution looks at only the third visit. Between email and display, email was first, so email gets 100% credit for the $50 purchase.
-
Using first touch and a visitor lookback window, attribution looks at all three visits. Paid search was first, so it gets 100% credit for the $50 purchase.
-
Using linear and a visit lookback window, credit is divided between email and display. Both of these channels each get $25 credit.
-
Using linear and a visitor lookback window, credit is divided between paid search, social, email, and display. Each channel gets $12.50 credit for this purchase.
-
Using J-shaped and a visitor lookback window, credit is divided between paid search, social, email, and display.
- 60% credit is given to display, for $30.
- 20% credit is given to paid search, for $10.
- The remaining 20% is divided between social and email, giving $5 to each.
-
Using Time Decay and a visitor lookback window, credit is divided between paid search, social, email, and display. Using the default 7-day half-life:
-
Gap of 0 days between display touch point and conversion.
2^(-0/7) = 1
-
Gap of 0 days between email touch point and conversion.
2^(-0/7) = 1
-
Gap of 6 days between social touch point and conversion.
2^(-6/7) = 0.552
-
Gap of 9 days between paid search touch point and conversion.
2^(-9/7) = 0.41
-
Normalizing these values results in the following:
- Display: 33.8%, getting $16.88
- Email: 33.8% getting $16.88
- Social: 18.6%, getting $9.32
- Paid Search: 13.8%, getting $6.92
-