How Schibsted's AI model helped boost subscription sales
Schibsted
Schibsted developed a machine learning model to serve personalized subscription-sales recommendations to anonymous, non-logged-in front-page readers, using first-party demographic data (age/gender predictions from advertising) and sales insights to generate real-time, on-demand recommendations rather than batch processing. Of 158 candidate data points tested, around a dozen were retained as model features. A/B tests showed a 75 percent increase in subscription sales from front-page articles compared to previous models. The system is integrated with Schibsted's Curate content recommendation platform, using Flyte for orchestration and AWS DJL for model inferencing, and the team is transitioning to Tecton as a managed feature store.
Overview
Schibsted developed a machine learning model to serve personalized subscription-sales recommendations to anonymous, non-logged-in front-page readers, using first-party demographic data (age/gender predictions from advertising) and sales insights to generate real-time, on-demand recommendations rather than batch processing. Of 158 candidate data points tested, around a dozen were retained as model features. A/B tests showed a 75 percent increase in subscription sales from front-page articles compared to previous models. The system is integrated with Schibsted's Curate content recommendation platform, using Flyte for orchestration and AWS DJL for model inferencing, and the team is transitioning to Tecton as a managed feature store.
The challenge
Schibsted needed to serve highly relevant subscription-sales recommendations to anonymous, non-logged-in front-page readers despite a lack of robust reading history and user data for these users, one of the industry's toughest personalisation challenges.
The solution
Schibsted developed a machine learning model that leverages first-party demographic data (age and gender predictions from its advertising business) and processed sales data to generate real-time, on-demand subscription-sales recommendations rather than relying on traditional batch processing. Of 158 data points tested, only about a dozen were retained as model features. The model is fully integrated with Schibsted's Curate content recommendation system, blending its insights with editorial signals, using orchestration tools like Flyte and AWS DJL (Deep Java Library) for efficient model inferencing in a Java-based internal application, and the team is now transitioning to Tecton, a managed feature store solution.
Reported business value
A/B tests showed a 75 percent increase in subscription sales from front-page articles compared to previous models, in some of the use cases, unlocking new revenue opportunities.
Sources
Open any source and check the claim yourself — that is the point of the register.
Other media & entertainment entries in the register.
How Business Insider's AI-based paywall strategy increased conversions by 75%
Business Insider replaced its editorially-driven freemium paywall with a machine-learning-based smart paywall that decides which content to paywall for each user based on prior reading habits, referral platform, and each content genre's propensity to convert. In testing from December to April, total conversions increased more than 75% versus the control, with 60% of new conversions coming from stories that would never have been paywalled before. Applying AI to the registration wall as well increased registrations by 300%.
Süddeutsche Zeitung: Lokalinformationen per WhatsApp
Süddeutsche Zeitung (SZ) launched WhatsApp channels for districts around Munich, offering free local news, event information and regional updates. Content comes from SZ's local newsrooms and is supplemented by automatically generated information produced in cooperation with Berlin media startup Beat Squares, which uses AI to process data from publicly accessible sources. All contributions are editorially reviewed before publication. More than 4,000 users subscribed within the first three days.
Wiley boosts service efficiency with Salesforce Agentforce
Publisher Wiley uses Salesforce Agentforce alongside Service Cloud and Einstein AI to manage customer service, letting AI agents resolve common issues like password resets and account access so human reps can focus on complex cases. Agentforce improved case resolution by over 40% compared to Wiley's previous chatbot, helped onboard seasonal agents 50% faster, and delivered a 213% return on investment with $230,000 in annual cost savings.
ElevenLabs: Scaling Voice AI with GPU-Accelerated Infrastructure
ElevenLabs runs its voice AI platform for text-to-speech, voice cloning and speech-to-speech translation on Google Kubernetes Engine (GKE) with NVIDIA GPUs, using GPU optimization strategies such as Multi-Instance GPUs and GPU time sharing to improve utilization and reduce costs. The platform generates 600 hours of audio for every hour of real time, supports 29 languages (expanding to 40), and is reportedly used by 41% of Fortune 500 companies, powering products including AI Dubbing Studio, long-form audio for audiobooks and podcasts, article narration for outlets like The New York Times, and real-time conversational AI assistants.
Was this helpful?
Your feedback helps us improve our use case database