Mapping the customer journey, tracking lifetime value, personalising offers, and balancing data and human judgment are the table stakes in subscription marketing analytics, said Professor Przemyslaw Jeziorski of the University of California/Berkeley in an interview with INMA.

As data became increasingly important to understand news media subscribers and improve profitability, Professor Jeziorski urged media companies to start small.

At a recent quarterly meet-up, he told executives of 160 news brands benchmarking with INMA: “Data collection is easy and cheap, so many companies build lakes and fill them with trash. The problem with this is that data retrieval and analysis is costly.”

He recommended marketers start with the data they already have and plan analyses based on the customer lifecycle, which includes an acquisition stage, development stage, and retention stage.

“While everybody is excited about data lakes, machine learning algorithms, and real-time analytics, you can actually do the key analyses in Excel.”

Map the customer journey

To effectively use data, advised Professor Jeziorski, it’s important to map the customer journey, understand how people make decisions at each stage, and start using data to tailor products and offers to individual customers.

“I would start with retention analytics and identifying customers at risk of leaving,” Professor Jeziorski said, explaining his reasoning: 

• Most likely you already have all the data you need, such as each subscriber’s transactional history and insight about her Web and app reading. 

• The mathematics of predicting who is at risk of churning is simple because there are only two possible outcomes: a reader retains or does not.

• The project is cheap because you can run a basic logistic regression algorithm even in Excel.

• The insight can be easily applied to business: Send a discounted retention offer to a subset of at-risk subscribers, and measure a difference in retention rate vs. a group that has not received a discount.

Track customer lifetime value

“As media companies change their business model from mass-distributed goods to services, they need to measure the value of individual customer relationships,” Professor Jeziorski said.

Lifetime value is a simple mathematical model that gives companies one number to summarise the profitability of a customer and compare it to acquisition costs. It takes into account things like churn rate.

“Let’s say your average customer lifetime value is lower than your acquisition cost. Why is it?” asked Professor Jeziorski as he considered possibilities: “Is it because you are spending too much on promotion, charging too little for the subscription, or people are churning too quickly?”

By calculating lifetime value, companies can do back-of-the-envelope scenarios to understand the potential impact of changes, such as a decrease in churn rate: “How much would lifetime value increase if we cut our churn rate by half?”

Lifetime value can be calculated for any customer, including old and new customers, and can be used to make decisions about investments in marketing, product, or data analytics. 

Personalise offers

In digital subscriptions, there are various paywall models that target potential subscribers based on criteria such as Web site usage or content type. However, targeting can also be based on individual motivations and preferences. 

“Let’s take a complex product such as The Wall Street Journal that delivers all sorts of benefits to different people. Some sign up for the latest news. Some others want to support quality journalism. Think of these two benefits as two products,” Professor Jeziorski said. 

Applying predictive modelling, we could not only decide whom to target with our acquisition or retention offer but also which benefit to highlight to whom. These offers can convert at higher rates than traditional one-size-fits-all messages.

“Another benefit of personalisation is reducing waste of customers’ attention. Although the monetary cost of sending e-mails is low, the mental cost for the recipients remains high, and people who receive irrelevant offers might ignore them altogether.”

Personalised prices can also be effective in increasing customer lifetime value. Many publishers differentiate introductory discounts or retention prices depending on customer behaviour (e.g., the lower engagement, the lower price). 

Professor Jeziorski warned publishers to consider the potential negative effects on customer trust and perceptions of fairness.

Balance data-driven vs. data-informed decisions

For Professor Jeziorski, data analytics is a support system rather than a decision maker. 

For example, when talking about reengagement campaigns for lapsed readers, Professor Jeziorski suggested using a combination of algorithmic recommendations and curated picks. 

Churn modelling often leads to engagement as the biggest driver for retention. The problem is we know little about disengaged readers, our data is scarce, so it is difficult for algorithms to recommend content most likely to get them back to reading.

“Use the limited data to choose the topic or the section the reader engaged with in the past, but ask editors to pick the best articles from those sections for the final recommendation,” proposed Professor Jeziorski.

“Data is helpful at making small decisions, but human judgement is necessary for strategic decisions and decisions where we face the unknown,” he added. 

Interested in more insights about subscription analytics? Join 160 brands from 33 countries that benchmark their subscription performance with INMA.

Greg’s Readers First newsletter is a public face of a revenue and media subscriptions initiative by INMA, outlined here. Subscribe here. 

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .