Schibsted shares 3 requirements for Big Data success

As the last speaker of the 2017 Big Data for Media Week conference in London, Torry Pedersen, head of publishing of Schibsted Norway and former CEO and chief editor of VG, discussed how data science, product development, and editorial must cooperate to implement Big Data solutions.

“It is the interdisciplinary thing that creates the magic,” said Pedersen, who allocated 15 minutes every day to gather all his employees at VG and speak to them. “I profoundly think that if we are to succeed in the digital era, we have to upgrade interdisciplinary and we have to work with the culture.”

According to Pedersen, the three essential requirements for an organisation working with Big Data are:

1. Cross-discipline cooperation.
2. Autonomous teams.
3. Short feedback loops. VG is one of the most important sites in Norway, beating Facebook and Google in terms of popularity. One example of Big Data application is a dashboard of popular news that is updated every 20 seconds. The interface is simple, easy to understand, and designed for the newsroom instead for the data scientists.

Another example was Engage, an app that is driving changes in the newsrooms with personalised performance dashboard for all of their journalists. The newsroom adoption of these tools shot up significantly from March 2015 to approximately 880,000 unique users in Feb 2017. The interdisciplinary approach has resulted in a design that suits the needs of the users, hence the high adoption rate.

Some of the data-driven technology in the newsroom in VG are:

1. Optimisation of the position of paid articles on the VG home site. This includes real-time benchmarking of paid article performance in respect to their current position and time in this position on VG.no site.
2. Conversion helper effects. This has increased focus on data-informed decisions, increasing both the conversion rate for VG+ and the absolute sale from the front. The democratization of sales and conversion data in the newsroom have given ownership to VG+.
3. A/B testing as a self-service solution. More than 1,300 A/B tests have been started by the newsroom (20 tests per day on average).
4. Topic modeling. VG uses topic models to find the themes (topics) associated with a doc based on textual content.
5. Hunting relevant premium articles.
6. Personalised editorial newsletters.
7. Data models that can predict actions. They build a prediction model to identify the news users who will most likely buy a subscription. They can thus target only users that will most likely respond to an incentive. This will lower the cost of acquisition.

Pedersen’s interdisciplinary philosophy also extends to the practice of data journalism at VG. When working on big investigative stories, the newsroom often brings journalists and data scientists together. In one of the award-winning exposés by VG, the data scientists are included in the bylines, as Pedersen recognised their contribution and that different departments are dependent on each other.

Serla Rusli is a business and financial journalism MA student at City, University of London. She can be reached serla.rusli@gmail.com.