Essays on measuring user engagement with online news.

This thesis presents research on measuring user engagement with online news. The work done contributes to the domain of searching for better proxies of explicit user evaluations that can be used to improve large-scale measurements of user engagement.

The text is a bundling of three papers, written as a part of the work done in the context of the PhD programme at the Faculty of Economics and Business at KU Leuven.

All papers use data from our own experiments with over 400 users who read the newspaper on tablets. We collected both implicit and explicit feedback about the quality of the user experience on different levels of the experience.

We approach measuring user engagement from a data science point of view. Each thesis chapter applies predictive analytics techniques to address some issue in the area of measuring user engagement with online news.

By reading through the thesis, the reader will come across different interesting prediction problems: from predicting satisfaction and word of mouth behavior on the level of one reading session by contrasting features based on surveys with features based on tablet interactions; to demonstrating the usefulness of fine-grained user interactions for detecting explicitly indicated engagement on one news article by comparing traditional features based on time spent with features based on fine-grained swipe interactions; to showing the value of social network analytics for improving preference predictions about which article a user is going to read next.

The work is relevant for practitioners as it presents a mixed-methods case study for applying predictive analytics to online service experiences in the publishing industry, showing to what extent logs of fine-grained user interactions can be leveraged.

The thesis contributes to the field of user engagement by looking for better proxies of explicit user evaluations that can be used to improve large-scale measurements of user engagement in the domain of online news.