Multi-Touch Attribution - Date Windows

Date Windows

Attribution with the ChannelMix Modeling Engine allows you to customize what data is included in your model through the use of adjustable date windows. These date windows come in two flavors: the target date window and the lookback window. 

Target Date Windows

The target date window selector is used to customize which dates are used to train the attribution models. When a date range is selected, only the paths that ended during the selected window will be included in the models. There are multiple preset windows to select as well as the ability to define a custom window.

As an example, imagine you decided to try out a new marketing strategy last month. In this case, you should use the Previous Month window to focus the model only on paths that ended sometime last month. This allows you to focus on the value of your channels since you’ve implemented your new strategy. 

In contrast, you can use the Lifetime window to get a holistic picture of how your marketing channels interact to produce target events. With this window, ChannelMix will train an attribution model using every customer path we have available. 

Lookback Windows

Lookback windows are used to target certain consideration periods. The window defines a number of days prior to the last interaction of a path. Only interactions from this path that occurred within the lookback window will be included in training the model.

If you’re using Rule-Based attribution, you only have one lookback window option: Last 30 Days. This is because, although the Google Analytics UI allows you to adjust the lookback window, the API only provides data with a 30 day lookback window.

On the other hand, if you’re using Data-Driven attribution, we are able to reconstruct paths with any length lookback window. By default, ChannelMix provides five lookback windows: Last 30, 60, 90, 180 days, and lifetime.

If your customers typically have a short consideration period, something like a couple of days, the Last 30 Days window is perfect for you. In this case, for each customer path we will check when the last interaction occurred. Then, we will remove any interactions from the path that occurred more than 30 days from the last interaction. This allows you to ignore interactions from months ago that probably aren’t contributing to the recent transaction. 

The Last 30 Days window can become a problem for you if your customers typically have a much longer consideration period. What happens if it usually takes someone months to come to a decision? In this case, you will want to choose Last 90 Days, Last 180 Days, or maybe even Lifetime to make sure that your customers entire journey is included when training the model.

How Target Date and Lookback Windows Work Together

A combination of the Target Date window and Lookback window defines the earliest date an interaction can occur and be included in the model. For example, if you are using Last 30 Days for both windows, the earliest interaction that is included in the model could be 60 days ago. That doesn’t mean there has to be an interaction 60 days ago, but that there definitely won’t be any interactions included from 61 days ago.

The Lookback window is tied to the date when a path converts. This means that if the earliest target event occurred 28 days ago, the earliest possible date that an interaction can occur will be 58 days ago. 

The Target Date and Lookback windows work together to allow you to apply business logic to choosing which interactions are included in your attribution model.


Was this article helpful?
0 out of 0 found this helpful
Have more questions? Submit a request



Please sign in to leave a comment.