How do we keep our readers engaged? Well, a logical starting point is to show them relevant articles. Unfortunately, at VK Media, we weren’t really doing that.
Our front page was working well, but the article recommendations shown below the article on the article pages were not generating much engagement. With more visits coming from search and social media, the situation was concerning.
So, what to do?
Initially, our thought was: Let’s build a continuous scroll function. Looking around, many other news outlets used this, so why not?
But when we started to investigate the technical implementation, it was clear that this would be a major development project that would take up a lot of time for other stuff that we really wanted to do. And we didn’t really know that building this continuous scroll would give us good results.
Had it been a couple of years earlier, we would have gone for it, built it, and hoped for the best. But in the last few years we’ve taken big steps to become a data-driven organisation.
So instead of “swinging for the fences,” we took a step back, analysed the situation, and chose a truly data-driven and iterative approach. We focused on small improvements and constant testing without knowing where the data would take us.
Steps to success
The first step was to clean up the article footer to minimise the “dead space” between the article and the recommendations, leading to a 35% uptick in CTR.
The next step was to build a system to easily manage the recommendations instead of the “hard-coded” solution we had earlier. Within our CMS, we made it possible to quickly set up different lists of recommendations based on the metadata from the article. Through constant A/B testing, we could tweak what recommendations worked best. Over a period of two months, we ran 15 A/B tests to see what was working and what could be improved.
We found recommendations that were subject-wise close to the main article performed much better. By being able to show articles that were closely related, we also completed another main objective of the project: to minimise the risk that our readers didn’t get “the full picture” of a story containing multiple articles.
We also tested whether an article list, an article teaser, or the start of a complete article worked best. We found the teaser version came out on top in most cases.
Small steps, big success
By choosing the iterative approach, we kept constant momentum in the project and achieved excellent results, increasing the CTR of the article recommendations on the article pages by 130% from the baseline.
Moreover, we started showing native ads among the recommendations, creating a new ad space that has performed equally with the native ad spots on the front page in terms of clicks.
From bad to really good, we increased the engagement of our readers by using a data-driven, iterative approach that kept the project moving forward. We achieved excellent results, at the same time saving a lot of time spent on development compared to a big “continuous-scroll” project.