Xavier - a personalization system
Overview of this campaign
A traditional tabloid news site has different needs than a revenue driven commercial departement. But at the end of the day they both want the same things:
1.To present the right message to the right user at the right time
2. To not waste valuable space on the wrong content
So due to a growing need for good data, privacy and GDPR, and a techno-optimistic organization, we took matters in our own hands and made Xavier.
In short; Xavier is a solution that allows us to personalize any part of our sites in real time, based on the user’s behaviour and guided by editorial or commercial needs. We track, store and process our own data, and have full control of the core parts of our systems. Showing the right content without wasting valuable space, means balancing a combination of editorial needs and automation, both working towards increasing user engagement.
The editorial needs are met by always letting the editors define what content Xavier can access. This ensures that Xavier never shows content the editors haven’t approved. Xavier’s job is to increase our KPIs by showing the best content from the safe pools to the right users at the right time. Here are three important capabilities:
1. Speed: It takes about 50 ms from a user visits our sites until the recommendation is presented for them. Allowing Xavier to personalize any position on our pages without diminishing the user experience.
2. The models: To meet different needs, Xavier can combine and AB-test various types of models.
3. The flexibility: Having full control over the tracking, the system and models means that we very quickly can change and adapt when new products are deployed or new models are needed.
Xavier is now present in one form or another on all of our websites.
Results for this campaign
The main asset of Xavier is the power of having full flexibility to change fast whenever that's needed. Different sites and products need different types of models and solutions, and Xavier has to evolve and develop to meet the different needs.
Here is some examples of how Xavier is used:
Recommendations under articles: All content presented under articles was already automated, but with no other logic than chronology. This became the natural first frontier for Xavier. A combination of models for collaborative filtering, and removing articles the user has seen or read outperformed the existing solution by 20 - 30 %, and Xavier are now recommending all content under articles.
Videos on Dagbladet: About 80 % of the video traffic from Dagbladets front page comes from Xaver, combining models for popularity and removing those the user already have seen or doesn't want to see.
Premium content on Dagbladet: Almost all of our premium content are recommendations from Xavier, by using a model for recommending articles that drives subscriptions we now recruiting several paying users without the need of manual input from the editors.
In addition there is implemented a solution for optimizing our Content Marketing campaigns. Xavier is also recommending a large portion of our free site articles on all our sites, using a combination of collaborative filtering or popularity, and removing unwanted or read articles.
The common denominator for all of these solutions is that they couldn’t have been implemented if they didn’t add value. And they do add value because they combine the need for editorial control with personalization and automation.