As Schibsted’s Siri Holstad Johannessen took the stage Thursday morning at the INMA Media Subscriptions Summit in London, the predictable “Hey, Siri!” and Alexa jokes might have been a tip-off to the almost sci-fi nature of the discussion to follow on the subject of using new systems and technology to drive news industry culture and growth.

Who knew that subscription conversion now resides in the realms of machine learning, random forest methodology, pattern recognition, data visualization, and Artificial Intelligence?
Well, for two: Johannessen, the Norwegian publisher’s head of sales and marketing, and Steven Neubauer, managing director of Switzerland-based NZZ.

“One key prerequisite to apply Artificial Intelligence or machine learning methodology to your business is a unified data warehouse,” Neubauer advised. “If you don’t have your data in one place so you can use it in those processes, if it’s not accessible. If it’s spread across systems and silos and functions, or different companies if you’re a large company, you won’t make progress there. You need to have it in one place. It’s an effort. It takes time. But you need to go through that process if you really want to have good results out of it.”

Neubauer also recommended having data science experts reside on the business side, not in the technology division where they won’t be immersed in a business mindset or see the immediate results of their efforts.

“Data is super important,” said Schibsted’s Johannessen. “And data is increasingly more used in Aftenposten, and we’re trying to use data for every decision we make. But we still consider it early days and we’re not done. We’re just seeing the start of how you can use data to grow subscription sales.

“I’m so glad to see that media houses are really focused on using data and visualising it and making it accessible to the whole organisation. Our tool, called Insight, is there for every Schibsted employee working with media. All the journalists, all the business side, can see this. And that really makes us move in the same direction, making good choices, working with optimisation in the right way.”

NZZ’s Neubauer said: “No. 1 for us is to build scalable platforms, central platforms, and that infrastructure that allows for experimentation, and to accelerate those learning cycles. In the last months, we’ve really exponentially increased the learning cycles that we’re going through. In the beginning we were making, I don’t know, maybe one or two A/B tests per month, if we were good. Now we’re doing dozens each day. It really helps to go out and test to see what’s working and what’s not.”

Johannessen emphasised that data collection is only the start.

“Once we have all this data, we need some analytics to make it even more useful,” she said. “We’re doing a lot of data modeling and predictive analytics support to support all our teams. Algorithms give us insights into patterns for those who convert and those who don’t — combining subscription data with visitor data or log-in data and telling us who we should target. It’s working really, really well.

“So, we are using algorithms now. And the data team has made a tool so that every time a subscriber calls the service centre, the subscriber will have a red or green or yellow light showing [customer service if the caller is] likely to churn or not. This is a very useful tool helping them to personalise the conversation. If [the caller is] not likely to churn, you can upsell or you can help them do other things. If they are likely to churn, OK then, you need to give them some kind of price reduction or a long subscription period or something like that.”

Neubauer agreed that such tools are the key to effective subscriber interactions and explained that NZZ introduced more flexible dimensions to its “paygate system” (NZZ eschews the term paywall). 

“We started to be able to change the call to action, the format, colour, personal greetings, price communication,” Neubauer explained. “We do it depending on the time of the day, the device, the reading behaviour, all those things. And we do it on the registration prompt, as well on the order prompt and also on landing pages. Starting to experiment with those dimensions, we have already more than doubled the conversion rate on the paygate.”

Having a concrete goal is important when experimenting with this type of technology, Johannessen said.

“We strongly believe that we need to change our mindsets and we need to work on retaining these customers that we actually have,” she explained. “I was talking about the 100,000 [new subscribers] goal. I think the next 100,000 digital subscribers will not come from sales. It’s going to come from keeping the customers we actually have.”

These responsive, data-driven technologies also present new challenges. And news media companies must be willing to experiment a lot, fail sometimes, and adapt quickly to what they learn.

“We did a lot of things wrong,” Neubauer said. “One, in the beginning, was not enough focus on core products. We also fell into the acceleration track — ‘Oh no! We need to be first mover on Instant Articles in Switzerland! And we need to be first-mover on the next thing, and, and, and, and ….’

“But in the end, when you look at what’s going to be the outcome of those initiatives … being the first-mover on those things didn’t really pay off. What matters is that you have an app and a Web site that’s scalable where you can run experiments, and that you improve for your subscribers in order that they stay with you.”

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