One year ago, we were facing a basic challenge well known to publishers around the world: When our users behave so differently from one another, how do we cater to all of their needs? We cannot continue to target an “average user” that does not exist.

Aftenposten users differ greatly in how often they visit our site and what type of content they consume on the site. Yet until recently, we had been serving the same one-size-fits-all Web site experience to everyone. Our “news junkies” were exposed to the same stories over and over, while less frequent users missed our best and most relevant stories because they could no longer be found on our mobile front page.

In other words, we were far away from a tailored news experience. Based on that realisation, we set out to build a more relevant and engaging front page that takes into account differences in user frequency and content preferences.

Aftenposten set out to create a front-page algorithm that also ensures content deemed journalistically important can rise to the top.
Aftenposten set out to create a front-page algorithm that also ensures content deemed journalistically important can rise to the top.

This could have been a straightforward algorithm optimised for engagement, but we wanted to do something more: We wanted to create a product run on sophisticated algorithms that would steer clear of so-called filter bubbles and echo chambers so often associated with Facebook and other algorithm-based news products.

Instead, we built a news algorithm based on our journalistic mission that actually closes the gap between what people know and what we think they should know. We call it the editorially guided algorithm.

We had four main objectives when we started on this journey:

  1. Create a more relevant and engaging product for our users.
  2. Build a product that better supports our business model of digital subscribers.
  3. Make sure more of our users are exposed to a higher number of our best and most important stories.
  4. Release a product that enables us to speed up product development and build completely new and enriching features.

One year later, the results have been impressive and the learnings plentiful:

  • We created a more relevant and engaging product for our users, illustrated by a significant lift in click rate from the front page.
  • We have built a product that better supports our business model for digital subscribers, illustrated by a significant lift in conversion rates on our new front page, and we believe this is just the beginning.
  • More of our users are exposed to a higher number of our best stories, as the new front page is built primarily on signals that reflect our editorial priority and time since published.
  • The new product enables us to speed up product development and build completely new and enriching features. By combining editorial judgment, relevant metadata, and sophisticated algorithms, the opportunities are almost endless.

Most importantly, we have demonstrated that it is possible to make use of the great advantages of data science, algorithms, and machine learning to offer a more individual news experience without filter bubbles and echo chambers. It is our responsibility as a trusted news brand to enlighten public debate and create the foundation for a better working democracy. We are doing that by combining sophisticated data and algorithms with editorial judgment and ethics.

In addition to all of the above, a data-driven front page enables us to spend less time on pixel-perfect editing, and instead focus on improving quality of our stories and distributing our content. Eliminating legacy workflows is a very positive secondary effect.