The emphasis in publishing media is currently shifting from a primarily ad revenue-driven model to a more audience-based one emphasising subscription pricing, so acquisition and price elasticity are key factors.

In addition — especially with a regional publication having a clearly circumscribed customer base — retention is an even more critical consideration in stemming the decline in subscriptions seen in recent years in the print media industry. Therefore, maximising customer engagement is key.

Understanding when customers might opt out of a subscription allows media companies to reach out to them at an appropriate time.
Understanding when customers might opt out of a subscription allows media companies to reach out to them at an appropriate time.

Certainly, in general, the way to do this is by “giving the customers what they want.” Ideally, we would tailor content specifically to each individual, but we’re not there yet. So we use analytics to identify clusters of subscribers/users with similar interests.

Also, what the customer wants probably includes getting all content for free, which is obviously not part of our model. Instead, we use analytics to optimise the subscription pricing, taking into account propensity to churn plus a variety of demographic and transaction history data.

Another aspect of the customer interaction that can be tailored is the communication strategy.

One way we use analytics in this context is optimising the timing of billing messages. Instead of one fixed schedule for reminder notices, we individualise the schedule so a subscriber who habitually pays on time will get a notice sooner than someone who pays three weeks late on average.

In this way, we look at a late payment as a deviation from the subscriber’s normal behaviour — and a signal that something may be wrong. And that’s not to mention avoiding making subscribers feel pestered, and perhaps becoming even more likely to stop their subscriptions.

We have tested this strategy and found it can provide a significant reduction in churn when implemented correctly.

So, what other signals can we look for indicating a subscriber may churn? Again, analytics provides an answer through the development of a propensity-to-churn model.

Naturally, we look to identify individuals who have a high score. However, just as with the payment deviation, it turns out that subs who show a sudden increase in their churn score (dubbed “accelerating churn”) are actually more at risk.

With this information, we can then institute a directed programme of intervening with subscribers who are likely to churn before it happens.

With this in mind, the question becomes what is the best way to intervene? We have been testing a variety of approaches to the high-churn and accelerating-churn subscribers. The fundamental measure is which approaches reduce stop rates the most versus no intervention.

But we also have to take into account whether a given intervention pays back, comparing the cost of the intervention against the lifetime value of subscribers who are saved. It turns out advanced analytics can further optimise the process through the development of a net lift, or persuasion model.

This type of model combines modeling with experimental design to systematically identify the characteristics of individuals who will respond best to each treatment, or none at all. We thereby address retention with optimal efficiency.

The final topic for today’s installment is matching content to subscriber interest. We did a content interest survey and compared the interest in various sections of the newspaper to the space currently devoted to each. We then used analytics to determine the value of the subscribers expressing interest in sections which we are currently under- or over-serving. This allows us to make rational decisions about how to redirect resources.

The application of analytics to satisfy customers is limited only by the imagination. I look forward to exploring these applications in more detail in future installments.