Creating Audience Segments

Audience segmentation is a widely used technique in marketing and advertising space. This approach puts individuals into several discrete group categories, making it easier for organizations to understand their intended target audiences. 

At Northwestern University Spiegel Research Center, I was involved in an audience segmentation project. This endeavor was part of our consulting project with the Institute for Nonprofit News (INN) to provide data-driven audience insights to their members. As many news organizations make sense of their target markets through 5-digit zip codes, we used zip codes as our base unit of analysis. While news organizations have some sense of who their audiences are, often these perceptions are gut-driven rather than data-driven. The purpose of this project was to generate audience segments with some foundational information about the populations in zip codes to either confirm their gut-driven perceptions or provide new data-driven insights.

To generate audience segments I used the K-means clustering technique to derive audience segments. This is an unsupervised machine-learning technique that takes several feature variables, finds underlying patterns among the variables, generates higher-order segments, and assigns these distinct segments to all observations (i.e., zip codes).


We used the following community characteristic variables as feature variables:


Population, population density, population growth, and income were log-transformed to address skewness, and all measures were standardized before conducting the cluster analysis. Three separate cluster analyses were conducted on ZIP codes for metropolitan, micropolitan, and small town/rural. As a result, we generated a total of 25 distinct audience segments (12 for metropolitan, 7 for micropolitan, and 6 for small town/rural).


Check out the segmentation at my online dashboard here!