The core objective of our audience engagement team is to find ways to make our subscribers use our product more. Of course, journalism is the essential foundation for success in this area. If the reporting isn’t relevant, we don’t have a chance in capturing the audience’s attention.

To ensure we deliver the best possible editorial contributions, NTM launched a data-driven strategy for our newsrooms. This includes providing prioritised topics, crystal-clear reporter checklists, and monthly data-driven evaluation talks with every reporter.

Of course, the product and user experience as a whole are also very important parts of getting the audience engaged. The way we distribute journalism affects the chances it has to be read or viewed.

Compare this with the printed edition: If the paperboy has been sloppy and your morning newspaper is wet when you take it to your breakfast table, you won’t be able to read it.

This is similar to digital distribution: If your Web site is just a fast flow of random news, the likelihood of getting specific attention on the journalism is not good.

NTM took on that challenge and made our first attempts for personalisation on our 18 Web sites around Sweden. We began with a small, specific part of the Web site with the goal of improving subscribers’ experiences of our news product. To ensure this is effective, we narrowed in on individuals with hyper-personalisation.

We created a new widget called “Recommended for You,” which has a prioritised position in the fourth-highest place on our front page. The goal of the new widget is to present the “right” content for the “right” reader. “We wanted to test hyper personalisation in real time, (and) not use pre-calculated lists,” said Simon Vancoillie, head of technical development at NTM.

We included headlines for five different published articles. To maximise attention, we decided to make a list with a number from one to five for each of the articles, giving the highest of them a larger picture.

Each of the articles presented in the widget is selected based on the historical reading pattern of the unique subscriber over the last two weeks. Based on the reader’s interest, articles are selected from a top list of articles in those topics, aggregated in real time. The top list is based on the article’s news value (ranked from one to five, based on an editor’s decision), how many subscriber pageviews the article has received, and how many comments and emojis/reactions it has recieved.

We do this to try to identify what kind of articles the subscriber would most likely engage in. When selecting the articles to be presented in the widget, we decided to exclude all articles the reader already read. This way, readers are only exposed to new stuff.

This is the “product” we released in November on our Web sites. After releasing, we continued to develop the functionality. We realised that if a subscriber didn’t click on any of the five articles, those articles could remain in the widget for a very long time, so we decided to make an automatic switch. If an article appears in the personalised widget five times without being clicked, it disappears and a new one appears. Implementing this change almost doubled the traffic from the widget to articles.

“The absolute hardest part was to create an infrastructure that is able to aggregate data for each individual and sort out the content the user has already read or has ignored, all of this in real time,” Vancoillie said. “We succeeded with centralising such a system and are now exposing APIs that all our products can use, including newsletters and push notifications.”

To do this, you need to have a business model that supports logged-in users and a professional data set up in place that gives you access to solid first-party data. All the things described in the personalisation process actually happen in real time very fast without slowing down the Web site.

“The results of this are so far pretty impressive,” Vancoillie said. “The system is also prepared for exposing different types of content, not only journalism, so we can use this commercially as well. For instance, displaying the right e-commerce products for the right user.”

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