Data used to personalise products, inform decision-making
World Congress Blog | 15 May 2019
Canada’s Torstar went from zero to data-obsessed in 45 days.
Croatia’s largest news publisher now runs on more than 50 data insight reports daily, weekly, and monthly.
Singapore Press Holdings built its own Artificial Intelligence (AI) recommendation engine to customise data-driven reader personalisation.
And Zürich-based NZZ found data success with an early reader registration gate in its conversion funnel.
For anyone seeking a touchstone Wednesday from the standing room-only Smart Data Summit at the INMA World Congress of News Media, there was likely at least one case study presented with which they could relate.
The Chartbeat-sponsored event at Thomson Reuters’ offices in New York spotlighted various insights data analysis can provide into growing reader loyalty and engagement.
“What we’re seeing in our datasets is indications of how to encourage each story to live its best life,” said Chartbeat’s Joshua Schwartz, chief of product, engineering, and data science at the news analytics firm. “We’re investigating the evolving way that news is discovered and how discovery affects whether readers subscribe.”
For example, according to Schwartz, whenever Facebook or any of the other major news-referring social platforms suffer an outage, rather than just not get any news until things are fixed, online users tend to switch to more direct sources.
“Rather than go down when Facebook was unavailable (for 45 minutes in August 2018), we actually saw a 2.3% increase in news consumption,” he said. “The moment Facebook was unavailable, we saw people switch over and start visiting publisher sites directly.”
Other data shows the emergence of a new category of news referrer undercutting the traditional dominance of Facebook.
“They are mobile-first aggregators, all of which have grown over two times in the past two years,” Schwartz said. “These are Google Chrome Suggestions, which is the suggestions you see on the Google search home page, and Google News and Flipboard. … These are sort of a mobile-first portal. These sorts of channels did not exist a few years ago.”
He also noted news consumers going to publishers direct from their mobile apps are nearly six times more loyal than visitors coming from platforms such as Facebook.
“Given all those sorts of indicators,” Schwartz said, “it makes one ask: Are we starting to be in a world where social is not the main way that people choose to interact with content, and if so, what comes next?”
The predominant indicator from the Summit’s four continents worth of case studies was that, whatever it is that comes next, it will be driven largely by intense data-derived personalisation.
John Souleles, chief data officer at Torstar, which includes the Toronto Star newspaper, stressed the importance of having an executive leadership team that believes in data as the company’s lifeblood.
“Don’t ever start with technology,” Souleles said. “Identify the use cases, build the strategy, the road map, hire the right people, then execute on the infrastructure. But key to all this is change management. … It’s not just that the CEO believes in it. It is the whole executive management team. That’s a big difference from a few years ago.”
With focused goals and a strong commitment, Souleles said, Torstar went from ideation to initial production of data insights in just six weeks.
And the company did it on its own.
“I’m not a big fan of hiring an outside company to come in and create a data solution for you,” Souleles said, “because if you have to rely on someone else to help you collect, access, and understand your own data, you’re not really a data company.”
Fiona Chan, head of media strategy and analytics at Singapore Press Holdings, said her company shared that in-house data development ethic when it came time to develop a recommendation engine for serving up content based on individual user preferences.
“How do we surface the articles that a particular user wants to read? If we do this well, what we’re trying to do is cater to a market of one,” she said.
SPH’s solution is a machine-learning effort to grade articles, and particularly its headlines, based on their emotional sentiment.
“We know that 80% of people read only headlines, and 60% of people share articles after reading only the headline,” she said.
In contrast, Boris Trupcevic’s goal as CEO of Croatia’s 24SATA has been to move his company away from emotional, “gut” decision-making.
“Don’t get me wrong,” he said. “I love the gut feeling. I’m a big believer still today. But that had to die in our organisation. … What I’ve learned from a lot of you guys is that, when you dive into data on subscriptions, you get really surprised at what turns people on.
“The more we learned, the deeper we dived into data. We started creating dozens and dozens of reports. We became really data driven in all our departments — editorial, advertising, subscriptions, circulation, everywhere.
“So, the end result of this process that lasted a couple of years is that, previously, we had a quarter of the people in the organisation using data. Today, with lots of training, we have 95% of people using data.”
Of course, any truly data-obsessed media house knows one of the first challenges is simply acquiring appropriate information to crunch for useful reader insights.
NZZ Mediengruppe tackled that issue by implementing an unusual early-stage registration gate on its Web site, explained Daniel Ammann, head of portfolio management for the Switzerland-based publisher. A show of audience hands at the Smart Data Summit found only a handful of other publishers use this tactic.
Ammann said NZZ has a strict meter of five articles before readers are asked to register for free, expanded access to the site: “Now we register about 10,000 new users every month.”
That registration information has become a treasure trove for the publisher’s follow-up efforts to personalise both the user experience and, eventually, subscription offers later in the conversion funnel. It has also allowed NZZ to diversify and expand its product offerings and pricing.
The result is a boost in total subscribers of 60% over last year — growth for the first time in several years.
“So our key message,” Ammann said, “was how important the registration gate was for us to identify users early on, how we personalised data products to increase relevance for and engagement of readers, how machine learning such as with the propensity score helps to get users to pay, and that the new product portfolio was important and worth the risk to succeed with digital subscriptions.”