First published on Digiday, November 2, 2023

For publishers, winning subscribers is only the start of the journey. Long-term revenue relies on moving from pure acquisition — getting as many conversions as possible and using aggressive trial offers — to building a sustainable, loyal subscriber base and maximizing customer lifetime value.  

Customer retention and engagement are crucial. As consumers double down on culling their subscription lists — 40% of consumers think they have too many subscriptions, according to a Kearney Consumer Institute Survey — publishers must prove the worth of their services over those of their competitors.  

As publishers focus on engaging and retaining subscribers and customers, they’re turning to personalization and propensity modeling, which leverages machine learning, to create targeted engagement and robust, long-lasting churn prevention strategies. 

Subscriber retention relies on the same personalization strategies that won the conversion 

To retain current subscribers, publishers are calling back to the techniques that won those conversions in the first place. They’re offering intriguing content that aligns with subscribers’ interests, segmenting audiences to provide the right content to the right people and engaging subscribers consistently.  

Piano’s benchmark data indicates that 68% of first-time website visitors don’t engage with the content. Additionally, new subscribers tend to cancel quickly by turning off auto-renew as soon as they convert. And yet those that do engage with content have the most significant potential to be lifetime customers, so engaging those audiences immediately keeps them close — and keeps them subscribed. 

“Publishers need to use a variety of tactics to encourage engagement and use of the subscription starting as soon as the user converts — signing users up for daily emails, encouraging app downloads, using browser and app notifications and creating content that drives habit,” said Michael Silberman, executive vice president of media strategy at Piano. “We recommend tactics that reinforce the value of the subscription by creating subscriber-only content or experiences and highlighting that for users through on-site labeling, email newsletters and content recommendations.”  

Predicting subscriber behavior with propensity modeling reduces churn 

With the help of machine learning, publishers are building propensity models to predict the likelihood of specific subscriber behaviors. For example, they’re leveraging data platforms to segment users as anonymous, registered or subscribed. This helps them see what topics, authors and formats attract, convert or retain each type of user — and which content prompts audiences to share their data. 

Additionally, these models can be tailored to predict a likelihood to subscribe or the likelihood that a subscriber will cancel within 30 days of the next renewal date, which is essential to strategizing customer retention. 

“If you’re just starting out with propensity, it’s prudent to use an existing model rather than trying to develop your own,” said Silberman. “If you’re a Piano client, deciding which model to use depends on your specific goals and use case. For example, if you’re trying to balance engagement and subscriptions, combining the return propensity model with the subscription one will give you segmentation by both loyalty and subscription likelihood so that you can target high loyalty/low subscription users differently than high loyalty/high subscription users. And, of course, test and learn, then continuously optimize and mature your approach based on proven outcomes.”  

Case study: How media companies dmgMedia and Fortune used propensity modeling to boost conversion rates and reduce churn 

The team at dmgMedia, a U.K. consumer media group with a range of global publishing brands, sought to capture its top-of-funnel audience and help them engage with the brand before presenting them with a paywall.  

To do this, the company partnered with Piano, using its Likelihood to Act propensity model to identify dmgMedia’s loyal users and show them the registration wall sooner. This tactic immediately boosted conversions.  

Next, dmgMedia used Piano’s Likelihood to Subscribe (LtS) model, which scored the probability of subscription action with a range of data points, including page views over the previous 30 days, desktop versus mobile pageviews, paid offers seen and active days with a paid offer. This provided the digital team with precise and actionable information. The conversion rate for the highest propensity segment was 174x greater than that of the lowest propensity segment — registered users were 30x more likely to subscribe than anonymous readers.  

In response, the dmgMedia team shifted its focus and engaged with this prime audience, leading to a five-fold increase in conversion rates for high-scoring LtS readers.  

Another prime case study example comes from Fortune Media, the American global business magazine. After building a robust subscription acquisition funnel, the media company turned to churn prevention, using Piano’s Likelihood to Cancel propensity model. The team leveraged the algorithm to sort its visitors into high- and low-cancellation risk buckets.  

This provided insight into where users were churning during the journey and also how they interacted with Fortune’s content up until that point. Gaining visibility into at-risk subscriptions allowed Fortune to create marketing campaigns to address these users, such as custom renewal terms.  

The right content, messaging and benefits remind users of the value of remaining subscribers 

Successful publishers are continuously providing users with the right content, messaging and benefits, utilizing many of the personalization strategies that secured subscribers in the first place. The goal is to gently remind each of why they converted.  

With machine learning propensity models, publishers are further identifying the best content for acquiring an audience and then what’s most effective for converting them — not just prompting the audience to engage but subscribe. Furthermore, using these models to predict subscriber behavior, publishers are determining who’s most likely to cancel and who’s most engaged so they can adjust interactions to increase loyalty and reduce churn.  

[:]