Introduction to Lookalike Models

Lookalike Models in itself contains groups of users that have a certain degree of similarity to the ones in the standard segment. User grouping is done using machine learning algorithms, based in their behaviour and the content they have consumed.

To provide the Lookalike Modeling feature, Bisko utilizes GjirafaTech's AIHub. The created groups can be used as segments to target specific users. Since these segments are created from Lookalike Modeling process they're called Lookalike Segments.

Lookalike Model Creation

A standard segment is needed to create a lookalike model.

After the seed segment is chosen, AIHub is instructed to train a lookalike model based on that seed segment. Users are groupd based on their percentage of similarity to the seed segment. After the model is trained, the user groups from the model can be used to create segments, namely lookalike segments.

Lookalike Segment Creation

When clicking a lookalike model that has already been trained, a screen as below will be shown, from where a Lookalike Segment can be created.

Segment Name is the name of the seed segment from which the lookalike model is trained and this field it not editable.

Reach and Similarity of the graph represent number of users in the group with x % similarity to the seed segment. You can choose any point in graph that isn't used until now to create a lookalike segment.