The Habit Project
Overview of this campaign
The Wall Street Journal’s Habit Project started with a lofty objective: to use machine learning and behavioral analytics to redefine how we engage with our members. Aggregating billions of member actions, we built, validated, and productionized a model signaling our members most at risk of churn. We then used survival analysis methods, borrowed from the medical sciences, to identify which habits adopted in onboarding best “extended the life” of a member.
Over two years of experimentation, our membership team developed a playbook of interventions to tactically increase engagement and reduce member churn. Our habit research drove the redesign of WSJ’s onboarding flow and member communications strategy. We refocused engagement marketing around the adoption of habits and highlighted “sticky” actions in onboarding. Throughout the customer lifecycle, we surgically applied our insights, targeting members that our churn model identified as having the highest risk of leaving.
Ultimately, we found that an engagement strategy focused on the formation of daily habits early in the life of a subscription was the key to reducing member churn. We coalesced around the goal of increasing members’ monthly “active days” in our digital products and developed a new analytics framework for measuring member engagement.
Results for this campaign
Since we started actioning the habit project, blended member engagement is up 3%, while member churn has decreased 3 ppt.—a significant win for the business.
We also now use our predictive churn model to surgically target our engagement communications [Exhibit 1]. To date, our member churn model has made 1.4 billion predictions on member’s likelihood to churn since it was productionized in April 2018. Within a 5% pt. margin of error, the model has accurately predicted churn rates for 98% of members. (For 60% of members, it predicted churn within a 0.5% pt. margin of error). These churn scores are appended to member profiles whenever they visit WSJ.com, so we can identify members with the highest risk of churn—based on engagement, tenure, and subscription type. We also use these scores to target engagement communications at members with the highest risk of churning.
While several initiatives drove notable results, some of the biggest lift came from driving incremental onboarding actions through a reinvented onboarding flow [Exhibit 2]. On average, new members now start 4 habits within onboarding, up from fewer than 1 when we began running experiments. Consistently, incremental actions in onboarding have driven stronger retention rates [Exhibit 3]. A single action can reduce churn by 5% over a member’s trial period, while members who perform three or more actions in onboarding have a 17% reduction in churn. By giving these sticky habits more visibility in onboarding, we’ve more than doubled adoption of several key habits [Exhibit 4]. We’ve also democratized this habit data, building an interactive engagement dashboard in Tableau so our membership and product teams can monitor habit growth weekly [Exhibit 5].
Externally, the Habit Project has been featured in Nieman Lab, Business Insider, and the Bottom Line Marketing Blog. We also found opportunities to share insights and best practices with local and international publishers, presenting findings at the Facebook Journalism Project and INMA’s master class on digital subscriber engagement.