This is like Netflix-style recommendations, but for news and media, where editors set the rules of the game and algorithms handle the heavy lifting of matching each reader with the most relevant stories and content.
Traditional ‘one front page for everyone’ wastes valuable audience attention and ad inventory. Schibsted’s system personalises content and news recommendations at scale, combining editorial judgment with machine learning to increase engagement, loyalty, and monetisation across its digital media properties.
Combination of proprietary first-party user behaviour data at large Scandinavian scale, editorial expertise embedded into algorithms, and a tuned recommendation pipeline integrated deeply into Schibsted’s products and ad stack.
Hybrid
Vector Search
High (Custom Models/Infra)
Real-time recommendation latency under high traffic, plus cost and complexity of maintaining accurate, up-to-date user profiles and item embeddings across many properties.
Early Majority
Unlike pure algorithmic feeds, Schibsted explicitly blends human editorial control with machine learning-based personalisation, optimising for both engagement and journalistic values rather than just clicks.