Lessons learnt

  • Digital Rights Management

All participants used the DRM portal, but this usage was quite low. We identified the reasons why participants did not value the DRM portal:

    • The price for the VoD items was too high. The price does not stand on its own; it has to be positioned against the value of the complete product. The next issues go into detail about the latter.
    • Due to a technical issue, the video quality was not as good as at the offices of NPO and PPG. It’s important to really be distinctive with higher quality items.
    • The content offering was too limited. Users are used to have a great choice of channels and series, so it’s important to have a great offer of content.
    • The DRM app was easy-to-use; other aspects such as visual attractiveness and available help information scored neutral. They were not bad, but they certainly did not stand out.

There is a case for offering VoD packages based on personal preferences, which is also what we are seeing in industry. Of three formulas, participants preferred a subscription based on the genres they usually watch: a subscription over buying per item because they did not want the hassle of paying every month, and based on the genres they usually watch because the find the personalization of such a bundle to offer great value.


  • In house Recommendations for HbbTV

Due to the low user participation it is difficult to provide strong conclusions about the evaluation of the recommender applications in the pilot. However, the earlier user research conducted in the households did provide us with an interesting approach to offer recommendations. In that study, we learned how to make the recommendations take into account contextual factors related to the viewing of content in the household, such as group composition (who is watching), genre (the type of content, duration), time-related aspects (weekend vs. week, schooldays vs. holidays, and time of day), and mood (state of mind). For more information, see http://link.springer.com/article/10.1007/s00779-015-0861-0.


After the conclusion that the recommender received insufficient data to enable personalised recommendation based on individual viewing statistics, NPO decided to merge the project with an ongoing research “Linear broadcast programming schemes provide valuable data for video recommendations” that’s part of a master thesis research by Joost Negenman NPO.


Recommendations presented by recommender systems began in ecommerce but are now increasingly important for television and entertainment. Broadcasters like the Dutch Public broadcast (NPO) need to redefine their content proposition. Broadcast models based on linear viewing schedules are time dependent and are part of daily routine, Cooper (2015) “schedule helps organise the viewing experience, building habit and loyalty”, their content item slots are limited and based on aggregated viewing behaviour for viewer guidance. On demand models are based video libraries with personalization and recommendations models to guide their profiled viewers. This research investigates how existing broadcaster data sets can provide valuable data for video recommendations. Historical linear broadcast schemes and audience measurements where mapped and extrapolated to NPO’s reference target audience groups in order to profile the non profiled viewers and avoid harsh privacy legislation. This resulted in a list of guidelines and a Proof of Concept (PoC) how to service both the reference target audiences and the individual viewer. This thesis can benefit broadcasters in developing their content strategy, making use of the existing broadcast datasets to design recommendation models.


  • HbbTV as a central interface for second screen competition

People greatly enjoyed the social gaming experience. The inclusion of a competitive element caused a compelling experience and participants also had the feeling there were more discussions and conversations using this format. The application creates real added value to linear TV. People did not feel that the second screen app distracted from the show. The majority of the participants was positive about the ease-of-use of the application but effective user interfaces are crucial. Even more so since people often have little time to provide answers. Also, timely notification that input will be required by the participants is important. Other conclusions we can draw are that Tablets are better for second screen applications because when people use laptops they report to more easily open other applications and be distracted from the show. Participants appreciated the presence of scores, they are essential for the social gaming experience.  And at least for participants alone in the room, it is interesting to provide comments by participants in other households. In this case, this means allowing people to view why they were for or against a certain statement in the show.