Slate publishes a wide variety of articles ranging from culture and sports to politics and the merits of goat spamming every day. We’ve always been able to sort the more widely read articles from the less popular fare, but we’ve never been able to tell whether the readers who come to us for judicial coverage are also interested in breaking news via Slatest or getting advice columns like Dear Prudence. So we wanted to find out who Slate’s largest and most distinct groups of readers are.
We approached this question by identifying groups of readers based on their subsection-level readership behavior using a technique called fuzzy C-means clustering. This type of data analysis is used to uncover complex groupings of behaviors and attributes. To do this, we gathered more than 70 pieces of information—such as the number of times each user read a different blog or writer—between March 2016 and March 2017 from a random sample of 200,000 Slate readers. We then examined those groups to identify which pieces their members are consuming.
But how do we cluster readership patterns? In the figure below, we map the hypothetical readership of two largely retired Slate rubrics: A Fine Whine against Blorple Falls. About half of readers fall into Cluster A; they read a lot of Fine Whine and very little Blorple Falls. Conversely, about half of readers—who read a lot of Blorple Falls and very little Fine Whine—fall into Cluster B.
Now, imagine we expanded this chart from two axes to 70 and plotted each user in our sample according to his or her readership with all the content areas we tracked. We did that by running a C-means fuzzy clustering algorithm that grouped observations so the center of each group is as far as possible from all the other groups while minimizing the distance from each observation to its group’s center. In other words, the process separates readers into groups based on commonalities in how they interact with Slate.
The two most distinct clusters, seen in the chart below, are Dear Prudence readers and Politics readers. The “News Hounds” on the left side love politics and current events writers but skip Dear Prudence and its current and former authors, Mallory Ortberg and Emily Yoffe. Conversely, the “Prudie Fans” on the right side can’t get enough of Slate’s advice column but pass over Slate’s politics coverage.
These initial findings echoed what we know from user testing, in-depth interviews, reader surveys, and other methods of analyzing traffic data. However, this analysis did not answer all of our questions about how readers cluster their content consumption. We also saw clusters that were either too small or too heterogeneous to draw any conclusions from.
These methods have helped us to better understand Slate readers and their patterns. Identifying and defining these clusters guides Slate’s commitment to provide more Dear Prudence—including expanding live chats, podcasts, and newsletters—while also catering to our News Hounds with a redesigned home page to make sure they have all the latest news. For everyone else, we’re continuing to identify and refine additional clusters of readers to better serve great Slate news and commentary in a more personalized, intuitive way.