News UK used personalized newsletters to cut subscriber churn in half
Over the past year, The Times of London and The Sunday Times have been using machine learning to better understand the wants and needs of its readers in an effort to stop them from unsubscribing.
News UK’s self-learning algorithm, internally dubbed James, serves subscribers the kind of content most relevant to them based on their reading patterns, at the time and frequency most suited to their habits. This algorithm will essentially act as a “digital butler,” serving up content to individuals through email newsletters.
As a result, 49% fewer people canceled their subscription compared to a control group. The publisher declined to share what its current churn rate is.
“The defining challenge that all media organizations face is how do you build relationships with someone you have never met, meaningfully, at cost and at scale,” said Mike Migliore, head of customer value, The Times and The Sunday Times. “Without the right tools, it’s just not possible. Now, engagement is the highest it’s ever been. We have new customers coming in every day, and James has been no small part of allowing us to do that.”
According to the publisher, subscribers create a billion data points every day across The Times’ platforms. The self-learning algorithm ingests these data points, including information on customers’ propensity to convert to subscribers, as well as churn propensity, to serve up the right content to the right individual at the right time.
For James, News UK took a representative sample of 300,000 subscribers and registered users, 117,000 of which were subscribers, across a range of age groups and sent them daily emails to fit their content habits. Generalizing what type of content retains people is hard when each person has different needs and habits.
“This customer might read three times a week for two hours, another might read multiple times a day for five minutes,” said Migliore. “Without AI, it’s not possible to have a deep understanding.”
Over the course of the year, it has been comparing the findings to a control group, it found that 70% of people interacted with the emails. Only 15% percent of people opted out of the emails, and these were already loyal customers who had developed their habits, said Migliore. “They didn’t need us getting in their way,” he said.
James was the most effective in keeping hold of customers who scored low on News UK’s bespoke customer engagement score, which calculates metrics like recency, frequency, interaction and articles read to give a propensity to churn.
The Times and The Sunday Times has 527,000 subscribers, of those 286,000 are digital-only. It also has 3.75 million registered-access users who can view two articles a week, and it’s building up its data prowess to understand the customer journeys at each stage. Acquiring new readers is part of the puzzle. But keeping existing ones from churning is critical, and it’s also cheaper. The longer someone is a subscriber the higher their lifetime customer value.
Publishers are increasingly turning attention to customer needs rather than content itself when it comes to what converts. Titles like The Wall Street Journal have dynamic paywalls that show conversion messages at the point when people are the most receptive, based on self-learning algorithms.
Nine people from News UK’s data science team and the publisher’s project partner, Belgian digital publishing company Twipe, were dedicated to developing James, with 20 people overall, including journalists and marketers, contributing. The project was also funded by Google’s Digital News Initiative, which contributed €1 million ($1.2 million).
“The key goal was to leverage AI and news especially is an interesting challenge,” said Danny Lein, CEO of Twipe. “News has a different relevance throughout the day. What can AI do to drive relevant interaction for readers?”
Publishers are working out how to balance personalization with editorial judgment. Subscribers pay for the editorial decision-making from The Times. Too many algorithms in news feeds have led to finger pointing of filter bubbles.
“We have no intention of going down the Facebook route,” said Nick Petrie, deputy head of digital, at the publisher. “This is marrying tech with our journalism.”
The goal is to develop James, expanding to other communications channels like push notifications and social media, plus incorporating it into News UK’s data management platform to deliver more effective advertising to customers. Beyond that, it plans to white-label the tool to other publishers in Europe. But there are also benefits in editorial commissioning too.
“In terms of the newsroom, James has leveraged a huge amount of data around how we treat individual subscribers and understand which type of articles resonate with the reader,” said Petrie. “We are more data-informed in what we are writing.”
More in Media
The publishers who attended DPS were focused on the potential upsides of applying the technology to their operations while guarding against the downsides.
Now that ChatGPT users can surf the internet for information, some publishers are reconsidering the weight of the issue.
As Meta makes celebrity-like chatbots, ChatGPT learned to “hear,” “see” and “speak” while Spotify is piloting AI-translated podcasts.