Using advanced Artificial Intelligence (AI) to boost digital reader engagement @ Ringier
Overview of this campaign
There is no one-size-fits-all approach to capture attention. The window of opportunity lies within 5 seconds. Therefore, we need in-depth understanding of our customers’ interests and needs at that moment. We believe that by understanding the semantics of what people read, we can deduce their interests, context and intention on a granular level. The evaluation of market products revealed no satisfactory solution to analyze unstructured journalistic texts in German language with Swiss nuances. Furthermore, with over 530 products we need a holistic and generic approach: a central data and technology platform containing two powerful elements - content profiling and interest profiling - to supercharge our personalization and achieve a maximum uplift for engagement and revenue.
Content Profiling - During the content profiling process, all contents are screened, enriched with public knowledge databases and analyzed using different advanced AI technologies. For written content, we use Natural Language Processing (NLP), Named Entity Recognition, Entity Linking and Sentiment Analysis. Pictures are screened with image processing technologies, e.g. Face Detection, Landmark Detection and Object Labeling. Then our self-build semantic engine “Ringier TagCloud” visualizes and clusters the contents after themes and entities (e.g. persons, places, events) using NLP and Machine Learning. Our solution not only showed greater precision, but also produced less noise than other market solutions. Additionally, our cloud-native architecture provides a cost-effective way to supply the computation power of 700.000 GPU. That way 1.2 Mio articles and images have been processed.
Interest Profiling – The combination of the data from our cross-portfolio user behavior tracking together with detailed information gained from the content profiling enables us to unveil hidden dependencies among users and build valuable user interest profiles using collaborative filtering. In advantage to market solutions, we build profiles based on three taxonomies to serve advertising (IAB) as well as media and news (IPTC) purposes. Our calculation of reader interest profiles happens real-time, allowing us to display the right content in the right moment.
Results for this campaign
Increased efficiency through automation: The knowledge generated by the “Ringier TagCloud” can automatically match appropriate images to contents, create dossiers about a specific subject and automatically index our content and websites for SEO on an extremely high quality level. These use cases reduce manual workload, increase quality and improve efficiency.
Recommendation and personalization boosting reader engagement and ad revenues: The “Ringier TagCloud” furthermore enables fully automated content recommendations by directly linking relevant and related up-to-date content to articles ("More on this subject”). These automatic article recommendations generate 32% higher Content Click-Rate than manually chosen recommendations from the editorial team.
Personalized content recommendation using the generated interest profiles for individualized content for each individual user on the homepage generated a 46% higher Content Click-Rate.
Personalized content recommendation using the generated interest profiles to address specific user segments results in a 127% higher click rate compared to non-segmented users.
Affiliate marketing generating new revenue stream: A cross-selling use case is “Affiliate Marketing”. We believe, based on the in-depth understanding of interests and context, we can derive and predict intent, for example whether a particular user will click a certain ad. Therefore, we identify users that convert exceptionally good on the eCommerce content as well as their look-alikes, and display personalized ads. We recorded a significant uplift of the Click-Through-Rate by 200%. This confirms that the segmentation of readers based on their interests can successfully drive conversions.
The selected and newly composed cutting-edge Artificial Intelligence technologies make this project unique and innovative. The platform improves the user experience by enabling several personalization and segmentation use cases. In consequence, the improved user experience leads to an increased usage of our sites and therefore to a better monetization. By explaining our data and technology platform and showcasing its usage, we hope to contribute to the media community with sharing this experience of leveraging technology as a vehicle for transformation.