ProSiebenSat.1 Digital Data cooperates with Berklee Valencia on data and AI topics for the second time
ProSiebenSat.1 Digital Data (short: PSDD) is a 100% subsidiary of ProSiebenSat.1 Media SE, which sets its focus on data-driven products and artificial Intelligence (AI). Berklee College of Music, Valencia Campus, is the European branch of Berklee University in Boston, USA.
Alexandre Perrin, Professor at Berklee Valencia, initiated the module “Data Analytics in the Music Industry” in the Master of Arts in Global Entertainment and Music Business. The following collaboration between Berklee Valencia and PSDD was led by Stefanie John, Associate Director Data Strategy, in the team of Fahim Alefi, VP Data Strategy. Stefanie John, who has worked in the entertainment and music business focusing on data-related topics since 2009, joined PSDD in 2019. Generating new impulses, the project is considered promising due to collaboration of digital natives with a leading entertainment house. From January through May 2021 graduate students in ‘Global Entertainment and Music Business’ worked on topics with a strong data and AI alignment attending the class ‘Data Analytics in the Music Industry’.
During research and empirical stages, the students tackled a problem and contributed to finding innovative solutions. PSDD selected the main topic ‘The Future of Media Consumption and Production’, but didn’t set too many boundaries to let the students pick their own focus area. At a second stage, the students created their own data/AI products considering technological developments of behavioral segmentation and media synthesis. In biweekly video calls PSDD provided guidance and continuously challenged interim results. The students delivered their final results using different methods packaged in a presentation.
Topic #1 was the future of media consumption in virtue of behavioral segmentation. The groups were able to select between subject 1 ‘Improving the User Journey of a Streaming Platform like Joyn’ and subject 2 ‘Optimization of User Engagement for an Audio-only Streaming Service like FYEO’.
The students firstly evaluated current trends on consumption partially caused by the pandemic situation. News was identified as major topic especially for podcasts [1]. Streaming services as Netflix and Spotify experience increased growth by nearly 90 million MAUS (monthly active users). Launched in 2020 Disney+ already has more than 100 million subscribers worldwide.[2] Podcast and audio book consumption increased by 7% comparing years 2019 and 2020.[3] Young adults are browsing 5h/day on average[4], 65% use their smartphone while watching TV.[5]
Based on these developments the following opportunities were identified. User generated content allows creating communities for niche marketing. Consumers themselves could be involved, e.g. in news reporting. Certain consumption situations require shorter as well as more concise content, whereas other situations might have a higher need for UX technologies such as AR and VR plus more interactive elements. Both therefore need to be identified, content needs to be adapted and integrated into targeting options. New target groups for podcast and VOD (Video-on-demand) could be children as well as older audiences. Last but certainly not least personalization is also highly relevant for news content since people have different interests and priorities based on their personal situation, mood, etc.
Topic #2 was the future of media production in the virtue of media synthesis. As the counterpart of the consumption topic students applied production trends also to streaming services such as Joyn and FYEO. Options for personalization and shorter content were evaluated from an automation point of view. In addition, they came up with auto-generated videos out of podcasts as well as the use of voice synthesis and even deep fakes, e.g. for news production. Also advertising could be adapted by adjusting advertising spots to the content consumed around it, e.g. changing the background music.
Having evaluated all ideas with business experts the following six seemed most promising and suitable for ProSiebenSat.1’s strategy:
- Music identification and live content streaming
- Master dashboard for streaming fans
- Clustering method for podcast users
- Adaptive advertising
- Audio-to-video transformation
- AI and deepface news production
- One-stop audio-shop
There were also other great ideas for general applications and digital products that will be forwarded to ProSiebenSat.1 relevant business owners.
1. Music Identification and Live Content Streaming
Integration of music on the platform allows firstly adding highly emotional content that secondly gives you further information on the users themselves. You can attract a new audience and differentiate from other streaming offers. With a deep integration of a music scanning software you can generate an additional experience, adding someone’s music library or enabling interfaces with other solutions extends the possibilities. Live content might be relevant also to linear programs, however to generate some added value for instance to premium subscription, exclusive live consumption could be very relevant for certain target groups.
2. Master Dashboard for Streaming Fans
Keeping an overview on streaming offers including a good recommendation still is a weak spot in the VOD world. If users had the option to connect different services, they would save time and the recommendations would profit from other services’ information. Users would provide their behavior by being more engaged as more traffic was generated on the platform. A product owner also would gained further insights into users’ interests and preferences.
3. Index for streaming users
In order to identify the potential for conversion into premium usage, a three-dimensional field is created consisting of a user score, an affinity score and streaming habits score. A low usage score can be boosted, e.g. by direct messaging and reminder pushes. By tracking and integrating individual information, content can be personalized to increase the affinity score. If the streaming habits score is too low, education and promotion of the service should be prioritized e.g. offering additional features like co-streaming parties.
4. Adaptive Advertising
Almost everyone knows the situation to be interrupted their on demand consumption by a completely unsuitable advertisement. An AI can avoid this by identifying the genre, nature and tone of the program and adapt the ad accordingly. Selecting the right background music or sound and changing even colors or wordings can improve the experience and put the advertiser’s message into a better light leading to more positive perceptions. This could be established as a new advertising format with different options of parameters to be adapted.
5. Audio-to-video format
If a podcast shows a potential for success, it might not be enough time and/or budget to invest into a high-end production that includes a RFP, casting of actors, etc. By reading text transcripts and other metadata of podcasts an AI could use visual templates to create a cartoon/video version. You would then have the option to also attract new video-first consumers, for example. At the same time, you had the opportunity to plan for a bigger production if also the video version goes viral. Also, this kind of production could be applied for a wider range of promising audio content.
6. AI and deep fakes
Especially for the news section being fast in scanning trends and producing in time is critical for success. And AI that scans relevant news, local ones or even from all around the world, translates them into other languages and enables a video performance could be a USP in the highly competitive news market. Text-to-video in combination with deep fakes could lead to an enormous growth of consumption as required technology improved enormously over the last years. To start with it could be launched as a podcast format that not only involves media synthesis but also behavioral segmentation.
7. One-stop audio-shop
Users are finally able to access all of their preferred audio content in one place with activity-based AI powered recommendations. After signing up, users connect and select their preferences. Their entire user journey, including discovery and continuous consumption, is tracked, so that an individualized dashboard can be generated. Podcast listeners can therefore be targeted based on behavioral segmentation integrating their favorite audio product into their daily habits.
This cooperation between PSDD and Berkley Valencia proved once again that our Company is focusing on highly relevant future topics. The focus areas and most ideas are very much in line with our existing innovation approaches. The students created also new impulses like the use of deep fakes, AI in advertising, connection of music & AI and how a personalized dashboard can improve your streaming experience. While behavioral segmentation is a topic PSDD is already working on, the students were also in touch with other business experts from innovation, technology and product ownership. The students appreciated working on research subjects facing opportunities and challenges of a leading European media company in 2021. They learned how to approach topics from a data strategy perspective as well as the role and power of data and AI in general. Firstly, they were put into the position of an external consultancy. Secondly, they were preparing a detailed briefing for a data scientist and/or engineer. Next year’s run could focus more on trend further AI topics, possibly with a focus on news as this becomes an even more important column in ProSiebenSat.1’s content portfolio. As new technologies and platforms quickly evolve in this highly dynamic, digital world, PSDD will introduce once again new cutting-edge topics of 2022. Going forward: Will a media company need to ‘collaborate’ even more with artificial voice technology and avatars? Can an AI dependably identify fake news? Which content will be adequate for which device?
To be continued.
[1] Seven.One Media GmbH; ID 1076171
[2] Q1/2020 to Q1/2021; Statista; ID: 196642
[3] Bitkom; ID: 876487
[4] Statista; ID: 253421
[5] Seven.One Media GmbH; ID:405361
