In today’s competitive and crowded news market, finding and retaining loyal readers is an essential component of news organisation sustainability. After all, with so many different ways to consume news, reader loyalty is hard to come by. The data team at the South China Morning Post recently set out to understand how readers develop loyalty to specific news outlets and how to nurture that loyalty.

In January 2019, the team began building an algorithm that uses machine learning to predict reader loyalty. We named our predictive engine Bluefin because we hope our readers will be drawn back to us to find informative, meaningful content, just as bluefin tuna always return to their birthplace to hatch. Also, both loyal readers and bluefin tuna, a critically endangered species, are increasingly rare. They both need protection and cultivation to return to a healthy population size.

While there are many potential applications to being able to predict reader loyalty, we are particularly interested in using it to optimise our marketing campaigns. We used A/B testing with a control group to check whether our predictive engine could improve SCMP’s marketing strategy. By focusing on the highest potential readers, we were able to increase engagement with our marketing campaigns, thus improving their cost efficiency.

We defined reader loyalty as when a multi-session user returned to the site with a pre-specified frequency and recency. Of the 40+ variables in the model, several stood out as top indicators of loyalty: percentage of pageviews in each section; time on page; duration between the last two visits; percentage of sessions on various platforms; and percentage of sessions by source and medium.

Using data from the period between June 2018 and November 2018, we built a workflow for the algorithm, which began with data extraction (refer to slider image 2).

The scoring engine measured the precision of the algorithm’s prediction of loyal readers. Each month, we feed the findings of the scoring engine back into the data engine to improve the model. This allows the algorithm to learn from the latest data and incorporate any new, relevant variables.

Recognising that the consumption patterns of SCMP readers vary from region to region, we ran the model separately for the US and Asia regions. We also used multi-month historical data to perform a cross-time validation on the model’s prediction.

In our multivariate A/B test campaigns in the US and Asia, we found that the application of predictive algorithms increased engagement by 58% to 78%, and cost-effectiveness by 36% to 52%. This proved that predictive algorithms such as Bluefin can optimise marketing campaigns effectively and efficiently. We are now applying Bluefin to identify new audiences who share the characteristics of existing high potential loyal users.

Identifying potential loyal readers is beneficial not only for maximising the use of marketing budgets and minimising reader churn, but also creating personalised reader experiences. Predictive algorithms, such as Bluefin, will enable marketers to pursue a much greater range of options which, based on a user’s preference and potential, should engage loyal readers in a meaningful way.

Bluefin’s success provided extremely valuable insight into SCMP’s readers. We trust that it will continue to be a powerful and dynamic tool, and we look forward to exploring iterative applications of Bluefin’s reader loyalty prediction in the near future.