Publisher Demo

Messaging customers with relevant information at the right time is one of the most important, albeit most difficult, aspects of marketing today. For example, sending an email blast to a group of users with the same creative, not knowing if those users are likely to purchase your product or service, is certainly not optimal. Instead, by using simMachines’s Dynamic Predictive Segmentation, marketers will be able to segment customers based on whether they are predicted to take a specific action or not, while using the dynamic weights that are driving each prediction. In this demonstration, we trained a classifier on whether or not a household is likely to purchase concert tickets. These classifications are represented in the “class” variable, taking the form “concert_positive” or “concert_negative”. Household demographic data, joined with publisher subscription data, was used as the input data for training this classifier. For demonstration sake, the subscription data used was more general than what one would see in practice.

The value of the “publisher_category” column could be one of the following:

  • Travel & Leisure
  • Automotive
  • Sports
  • Cooking
  • Lifestyle
  • Entertainment
  • Home Improvement
  • News

Let’s begin the demonstration by focusing on a segment of predicted concert ticket buyers in this visualization.

The innermost ring of the visualization represents the class variable. In this case, that is concert_positive or concert_negative. Fanning out from the class variable you can see the weighted factors for the selected segment. As you move out from the inside, the factor weights will decrease. For this particular segment of likely concert ticket buyers, you can see from the top factors in terms of their predictive weight that they share interests in books, magazines, and music. In particular, this segment subscribing to “entertainment” publications appears to be very predictive of purchasing concert tickets. Combining this information with the other demographic features found in this segment, you can see how a specific audience can be targeted by a personalized creative for concert tickets.

Conversely, we can also look at segments who likely will not purchase concert tickets:

In this segment, it is apparent that subscribing to “cooking” publications, especially in the geographic regions showing up as top predictive factors, is important in classifying these households as non-buyers of concert tickets. Sometimes knowing who and where not to market is just as important for marketers as the opposite.

This is the type of customer targeting that will need to be adopted to stay ahead of the innovation curve for marketers. Explore the remaining clusters in the visualization to have a better understanding of the granularity and predictive power offered by simMachines’s Dynamic Predictive Segmentation solution.

Learn more about working with simMachines