22.06.17

Press Release

Keyrus helps 20 Minutes create a content recommendation engine and optimize customer insight through its Data Science Boot Camp

  • #Data Intelligence
  • #Media
  • #Big Data Analytics

Keyrus announces that it has been helping 20 Minutes utilize its data so as to create a content recommendation engine and optimize its customer insight.

Utilizing the data to gain greater insight into the reader

In 2015, 20 Minutes decided to create a Data entity to utilize the information collected on its various content distribution sites and applications. Among its initial data-related projects, the Group wishes to test making personal recommendations of articles according to visitors' profiles. The objective is to keep the reader on the site for longer by proposing to them content that will interest them and so make it possible to gain their loyalty.

"On a site like that of 20 Minutes, visitors do not identify themselves", explains Winoc Coppens, Director of Information Systems at 20 Minutes. "Personalization tactics must therefore rely on cookies so as to capture customer behavior data and deduce from them what would be the most relevant information to push towards those customers on their next visit."  To utilize this volume of data, Data Science proves indispensable, but for that, individuals with very specific technical profiles are required.

A Big Data Boot Camp combining agility, expertise, and innovation

In 2016, 20 Minutes thus approached Keyrus, who suggested it participate in its Data Science Boot Camp. This initiative consists of proposing to enterprises with innovative projects that they work with future engineers trained by Keyrus in the latest Data Science technologies and supervised by Keyrus experts throughout the mission.

"We immediately signed up for the project, because it allowed us to work with individuals with a young profile, close to that of our target readers, and who provided us with an interesting outsider's view, at a cost that was reasonable for this type of expertise in Data Science", explains Winoc Coppens.

The stages for refining the Data

During the first stage of the project, the Keyrus experts brought to light the need to improve the precision of the data, so as to enable Data Science to be applied to them in a relevant way. "Thanks to the iterative cycles of the project, conducted using the Scrum method, we were quickly able to improve the quality of the data by adding the necessary tags on our sites and applications", specifies Sébastien Kaczor, Data Analyst in charge of the Data entity at 20 Minutes.

At the end of this first phase, thought was given to defining the best rules to be applied. This thought process was undertaken through workshops in which notably the Marketing team participated. Upon each proposal of content, an A/B Test was triggered to assess the appropriateness of the recommendation.

Validating hypotheses and obtaining new insights

In the space of just a few months, the project enabled 20 Minutes to significantly advance its thinking on valorizing its data. "By utilizing the data, we were able to confirm hypotheses we had", observes Winoc Coppens. "On each project, we are now going to utilize the data to validate our choices and provide new ideas."

As regards content recommendations, 20 Minutes wishes to continue with the tests, notably on mobile applications, on which there is a greater level of engagement by users. This initial experimentation opens up prospects for the integration of content recommendation in other setups, such as the Messenger bot 20 Minutes will soon launch.

"Keyrus's assistance in the framework of its Boot Camp entirely met our expectations and enabled us to swiftly progress in terms of enhancing our customer insight", concludes Winoc Coppens. "We were so satisfied with the content recommendation engine, that we have decided to adapt it at the next Keyrus Boot Camp 2017 to allow us to incorporate it into our new back office for journalists."

"We are very pleased to see the trust 20 Minutes placed in us in the context of our Big Data Boot Camp", declares Nicolas Marivin, Big Data & Analytics Manager at Keyrus. "This innovative project enables us to valorize our skills and fill out our teams with rare profiles of individual, directly faced with strategic business challenges."

Download PDF