How should personalization work in journalism when trust, fairness, and editorial responsibility are at stake?
In this episode of Insight Shapes News by Kilkaya, host Nils Ove Håland Riise speaks with Christoph Schmitz, Product Manager at Schibsted Media and lead of Curate, the content recommendation system used across several of the largest news brands in the Nordics.
Christoph has worked with content recommendation for a long time. In this conversation, he explains how personalization is applied in practice inside a large media organization, and where the real risks appear when editorial judgment, algorithms, and audience behavior intersect.
The discussion focuses on why segmentation models borrowed from advertising often fall short in a journalistic context, how dominant audience groups can shape recommendations in unintended ways, and why adding more data does not automatically lead to better outcomes. Christoph also shares how collaborative filtering is used at Schibsted, why it relies on reading behavior rather than tags or assumptions, and what its limitations are.
A central theme throughout the episode is the role of the homepage. We explore why readers see different stories, why this can be difficult to explain internally, and why the very top of the page still carries editorial meaning even as personalization expands elsewhere. Christoph introduces the idea of the homepage as a shared reference point — a place where readers understand what matters at a given moment, whether they read every story or not.
🎧 Topics we cover:
- What personalization means in an editorial context
- Why segmentation introduces structural bias in news recommendation
- How dominant user groups influence what journalism gets visibility
- Collaborative filtering explained through real newsroom examples
- How demographic data can reduce model performance
- Personalization as a tool for fairer distribution of journalism
- Why readers see different versions of the homepage
- The editorial role of the homepage top
- Why news products should not be designed like social platforms
- The “bonfire” metaphor and its relevance for shared public understanding
- Why personalization requires gradual, incremental change
If you work with personalization, content recommendation, homepage strategy, or editorial-product collaboration, this episode offers a detailed and experience-based look at how these systems function in reality, and what needs to be handled with care.
🎧 Subscribe to Insight Shapes News for conversations on journalism, personalization, and the strategic choices shaping the news industry.
Chapters
00:00 The Risks of Personalization
02:47 Understanding Personalization in Media
05:32 The Evolution of Personalization Models
08:10 Collaborative Filtering and Its Impact
10:54 The Role of Demographics in Personalization
13:35 Challenges in Personalization for Newsrooms
16:25 The Importance of Content Pool for Personalization
18:41 The Misconceptions of Personalization
21:17 The Future of Personalization in Journalism
24:22 The Role of Desk Editors in Automated Systems
26:59 The Concept of Liquid Content
29:32 Building a Personalization Team
32:04 Measuring the Value of Journalism
35:06 Visualizing a Personalized Homepage
Research paper on personalization and long-term reader engagement:
https://arxiv.org/abs/2510.09136
INMA case study:
https://www.inma.org/blogs/ideas/post.cfm/personalising-the-front-page-isn-t-just-smart-it-s-fair
Trailer photo credit: Foto: Mathias Broe.
