In the media world, the same question has been asked for decades. How do you deliver the right content to the right reader at the right time? Social networks found the answer. Media websites lagged behind for a long time. We developed our own personalisation technology. Thanks to our work with Recombee, we have taken it to a new level.
A homepage can't expand. Unlike social networks, only a limited number of articles fit on it. Personalisation pushes that limit further, because it tailors the selection to what we know about the reader.
In 2019 we launched our own recommendation system. It worked well, but we wanted more. So we switched to Recombee, a Czech AI tool that adjusts content in real time based on each user's behaviour. Today it runs on virtually all of our websites, including Blesk.
Recombee works across platforms, from the web to apps. It combines our editorial rules with the outputs of advanced machine learning. It does not only select content, it also keeps learning from how people react to it.
What was key for us? Transparency and control. Recombee is not a closed system. It is a tool in which we know exactly what we recommend and why. It also has excellent support and has proved itself on global platforms too. The result is clear: better tools for editors and a better reader experience.
What exactly does Recombee do?
Recombee tracks what is resonating right now and lets us respond immediately. On the homepage, we can quickly offer content that is current, attractive and has a high chance of actually being opened by the reader. The result is a better user experience and stronger advertising performance.
At the same time, it is not only about trends. Recombee also personalises content based on the behaviour of individual readers, so everyone sees what interests them most, regardless of what is trending. The homepage becomes a space that makes sense for a specific person. And that is exactly why readers come back to us.
What is next?
We will soon roll out personalisation on iSport too. Sport is a demanding environment for recommendations. Every reader follows something different. One follows five competitions, another only one club, and the rest of sport leaves them cold. Seasonality is also a factor. A fan who ignores hockey all year wants to know everything during the World Championship, from the results right down to Nečas's lost kit.
That is why we are planning two-level recommendations. First, the sports that interest the reader are displayed. Then, within those, the articles most relevant to them.
Táňa Lančová, by the way a former CNC colleague, is involved in the Recombee collaboration. We also use AI every day beyond content recommendations. Recombee is just a small part of what we can do. AI helps us automate work across all formats including audio. Thanks to our internal platform, CNC GPT gives each of us an assistant literally at our fingertips. It is not just about technology. It is about giving readers a better experience with our content, and helping editorial teams work smarter and more efficiently.
Thanks go to everyone who contributed to the project! Ilia Atlasov handles the setup and adjustment of scenarios, which is why the whole system is fast and easy to manage. Bára Drozdová from the BI team continuously evaluates the results and the impact of personalisation. That way we know what works, what no longer does, and how readers' interests change across our websites.
Comments