Here are some common FAQs, best practices and tips for building Look-A-Like Models. Please refer to this Lookalike Modeling Training and Methodology for more information on how models works.

• Q. How are the models created?
  
  • A. Permutive uses machine learning to find other users with common features to the seed cohort to generate a lookalike model.
• Q. How often do models get trained with new information. 
  • A. When you initially set a model to train, it can up to 24 hours for the model to gather information and train your model. From there, models train with new information every week, once a week. 
• How many users does a seed cohort need to train a model. 
  • A. Best practice is at least 5K users in your seed cohort, OR an influx of 1K users within the last day. The best way to determine this is to use your unique user count under the Insights  Audience Engagement Tab. 
• Q. What is the look back period of information a model uses to train. 
  • A. Models look at users who have fallen into the chosen cohorts used to train models within the last day. If the amount of users within the last day is insufficent to train, the model will not be able to complete. 
• Q. Is there a minimum amount of cohorts needed to train a model? 
  • A. Best practice is to choose a minimum of 5 cohorts when building a model. This will allow the seed cohort to have enough information to gather when building out a model. 
• Q. How many models can I have available within the dashboard at once. 
  • A. Unless contracted otherwise, the models limit is set at 250. 
    
    To get more information on how models are created. please click here.