Matthew Rowe and Harith Alani won the Best Paper Award at the Web Science Conference 2011, located in Koblenz, Germany, for their paper in collaboration with Marcel Karnstedt, Jeffrey Chan and Conor Hayes from National University of Ireland, Galway titled ‘The Effect of User Features on Churn in Social Networks’.
The paper explored how ‘churn’ in social networks can be predicted by assessing patterns in user behaviour within a community. Churn is the notion of the loss of users, a term borrowed from the telecommunications domain. A risk posed to online community managers and hosts is mass churning, where many users leave the community, and as a result the value of the community diminishes. Prior to this work there was no clear understanding of what behaviour leads a user to churn, thus rendering its prediction limited.
In the paper the authors presented a detailed correlation analysis of the relationship between behaviour patterns exhibited by users and their churn probabilities. The findings from such analyses over community message board data found that globally, users who participate more are less likely to leave. However distinct forums exhibited idiosyncratic behaviour where, in certain cases, the more central a user was to a network – i.e. the more information that flows through the user – then the more likely they are to leave the community, while for other forums the converse was the case. The paper also found significant network effects on a user’s likelihood to churn, where an increase in the number of neighbours or friends that leave a community influences the user in leaving also.
Such findings empower community managers and hosts in detecting the early signs of churning behaviour and allow them to act accordingly to retain users and maintain a valuable community.
Related Links:
