Modeling Explained

In this article

A review of the Modeling tab in the Permutive platform

Lookalike Modeling is a method of audience expansion that uses machine learning to identify users on your site with similar behaviors to your cohort of choice. 

Click here to read more about Modeling with Enterprise Workspace.

Tip: Modeling can only be applied to audiences built with first-party data. If a cohort contains third-party data, you will not be given the option to build a model from of it.

Model examples

Example 1 Example 2
If you don't produce much sports content but want to sell an audience of 'Sports Fans', you could build a small audience of users who read sports content on your site and create a model off of it to identify other users who are also likely to be 'Sports Fans'.

Please see our doc on setting up a lookalike model for a step-by-step walkthrough on building a model.

How to read the model

Once the model has been created, a graph will be returned with lookalike audiences you can create at different reach/similarities. Every point on the graph is an audience that can be automatically created by clicking on the point.

If you hover over any of the points you will see the size of the audience and how similar it is to your seed cohort (the cohort you based the model off of):


Weighting Explained

To see which audiences influenced the model, see the list of weights below the graph:


High positive weight
Indicates that users in the cohort are very similar to users in the seed cohort.
High negative weight
Means users in the cohort are likely not to be similar to the seed cohort.
Closer to zero
Indicate that these cohort don't significantly influence the model on their own.


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