Audience Data Mining Case Study: PBS & LunaMetrics

A Modern Love Story… A Data Mining Case Study for the 2018 Data Lover

Google Analytics 360 can be used to collect and process a wealth of data, and there are many opportunities to make use of it. But some companies want to take advantage of the powerful data mining tools offered by the Google Cloud Platform: enter the Google Analytics 360 export to BigQuery. Today we’re publishing a data mining case study developed by LunaMetrics and PBS, showing how Google Analytics 360 and the Google Cloud Platform were used  to classify audiences to improve user experience design, personalization, and targeting for marketing and messaging.

PBS television programming reaches millions of people, and its website PBS.org is an online content hub that supports that television experience and provides online video streaming content. PBS.org, like many websites, strives to understand its users and their needs for features and content by developing personas and audience segmentation. Personas often begin with anecdotal knowledge of customers or users and can be informed by many kinds of data, including interviews and other qualitative ethnographic data as well as surveys and other quantitative market research.

PBS was able to develop an additional approach with Google Analytics 360 and its BigQuery export: employing a data-driven method to classify audiences. PBS already had a robust Google Analytics implementation, with the default information enhanced by Event Tracking for on-page interactions and a wealth of internal information surfaced and stored in Custom Dimensions.

Data Mining Case Study Statistics

A data mining algorithm classified clusters of similar users based on a number of behavioral factors.

PBS partnered with LunaMetrics on a Data Science Solutions project to distill large and complex datasets like these into concrete, usable results and provide us with the content for this data mining case study. LunaMetrics applied data mining techniques to find patterns of audiences based on their website behavior. Using BigQuery along with Google Cloud Platform products such as Cloud Datalab and Cloud Storage, they were able to extract answers from over 330 million website sessions.

The analysis identified six distinct groups of users, for instance those who primarily focus on either particular kinds of content (such as news or information for parents) or features (with different preferences for watching video online or on TV-connected devices). PBS was able to use these findings to reinforce and refine their existing personas, now based on behavioral data.  Moving forward, these personas can inform the creation of new audiences to be used in remarketing, advanced reporting and content experimentation.

For more information, check out the Data Mining Case Study. For the technical details, check out Audience Modeling with Analytics 360 and Google Cloud Platform on the LunaMetrics blog.

Posted by Jonathan Weber (Lunametrics) and Daniel Waisberg (Google)