Enriching the user experience with a digital recommendation system

Matej, a digital recommendation system developed by the Czech-based company LMC and implemented in Finland, has significantly increased the number of page views for Etuovi.com and Iltalehti.

Recommending similar stories and content to the reader has long been a familiar practice for many online medias and digital services such as online stores. However, a skillfully crafted and implemented recommendation system can take a service’s traffic to a whole new level.

Matej has been adopted by Etuovi.com, Iltalehti and, subsequently, Kauppalehti as well. At its best, it has increased the number of recommendation-based page views by tens of percents

– Etuovi.com has been using Matej for a couple of years now. From the very beginning we noticed that page views coming from the front page increased by over 50 percent thanks to Matej, says Alma Media’s CTO Simo Syrjänen.

In practice, Matej highlights more relevant content for the users browsing Etuovi.com. When a user selects a specific listing on the website, the recommendation system brings up new listing recommendations that are increasingly tailored to their needs. On the Iltalehti and Kauppalehti websites, the recommendation system suggests content based on the user’s reading preferences, i.e., stories read by other users who have also read the same stories as them.

Syrjänen believes that the reason for Matej’s efficiency lies in its long development history. 

– It’s great that the system is able to make completely anonymous recommendations. That requires advanced algorithms and considerable calculating power. Scalability is important for us because millions of Iltalehti stories are read everyday. An efficient recommendation system increases this number significantly, he explains.

Ondřej Mysliveček, CTO of Alma Media’s Czech-based subsidiary LMC and the person responsible for Matej’s development, echoes his words. 

– Matej has been under development for years, so we’ve had time to make significant improvements to it. The key in the system’s efficiency is the fact it is not based on rules defined by the programmers, but in Matej’s ”capability” of observing how users behave and learning patterns by its own means (machine learning). And thus Matej can learn different rules and patterns on the same kind of service in different countries if people behave differently, he says.

Matej makes content easy to discover.

The goals of Alma Media’s digital development include developing media and services to better offer readers the right content at the right time, improving the user experience and providing advertiser clients with a relevant audience.

According to Mysliveček, Matej works with practically any service that features a lot of content that may go undiscovered by users. Without Matej users of the service has to search through the content and thus they need to know what they are looking for. But Matej can change searching into discovering, because Matej can offer content the user doesn’t even know exists and maybe never thought it can be of interest to them.

– What’s interesting about Matej and its algorithms is that every business is different. For example on job search websites, users usually only click on each job advertisement once or twice. We had to adjust the algorithm for Etuovi.com because one user can view the same listing quite a few times, he says.

Implementing the system is not difficult, but it requires some cooperation with the owner of the platform. According to Simo Syrjänen, the implementation has gone smoothly.

– Since Matej has already been harnessed to utilise Alma Media’s data once, it will be easy to implement within the same company in the future. Of course, the implementation is made easier by Matej’s professional product organisation. The service must be of the highest quality in order for us to use it, Syrjänen says. 

Syrjänen and Mysliveček believe that there will continue to be a demand for Matej in the future, in Alma Media’s Finnish operation as well as abroad. LMC is currently looking into ways to apply the recommendation system to food delivery services. 

– Many people always order food from roughly five of their favourite restaurants. We are trying to figure out if Matej could be used to recommend relevant alternatives to them, Mysliveček says.

Mysliveček believes that the success of the service stems from the fact that many people would ultimately prefer to keep browsing the website they are already on instead of using a search engine to look up additional information on the topic. 

– Above all, the recommendation system can make things easier for users. Ultimately, few people are willing to spend time looking up information using a search engine when there are easier avenues for finding relevant content, he says. 

Share article