Wantent and emotion analysis of audiences

The challenge: how can we measure audience engagement in a remote and scalable way?

As part of STADIEM's acceleration programme, VRT collaborated with Ukrainian AI-analytics start-up Wantent to measure audience engagement with the programme 99 to Beat. As the programme was adjusted from short-format (5min) to long-format (50min), we aimed to investigate how the structure of longer episodes, the narrative development, and the games themselves were experienced by a test audience. Traditional evaluation methods, however, like focus groups, are not cost-efficient, scalable, and trustworthy enough. This is where Wantent offered a solution.

The case

Before analysing 99 To Beat, the parameters of the test audience (age, gender, etc.) and the measurement criteria of the content (engagement, emotions, empathy, and attention) were identified by Wantent and VRT’s own audience research department.

The research involved 40 participants that watched the first, second, and tenth episodes, as well as a trailer for the new season of the game show and six short episodes from the previous season, using the Wantent platform and a webcam.

The AI-model analysed the recorded reaction videos. This model recognises the reactions and gaze (eye-tracking) of different respondents.

Finally, the AI model scores the reactions and provides analyses in PDF format, supplemented with graphs, tables, and heatmaps.

Lesha Shaldenko, Founder of Wantent: “We usually work with short-form content such as trailers and advertisements. To successfully execute this project, we developed an adapted platform and created a module that splits long-format videos into scenes.

After a series of studies and evaluations of the audience reactions, we provided VRT with a scoring system and recommendations to enhance the engagement and effectiveness of the show."