HRADIO pilot 1: results on the "recommendations" feature

A third feature explored during the HRADIO pilot 1 is the “recommendations” feature. This feature includes a recommendation system for radio stations based on publicly available radio station and programme metadata. Moreover, it also offers services based on the user’s current station. In this blog post, you can find an overview on how listeners experienced this feature.

Mock-up of recommendations feature

The recommendations feature consists of a web-based user interface. The user can select a service , as shown on the visual below in the upper-left corner. Then a list of recommendations is shown, together with a planet visualization of the same result:

List of recommendations and planet visualization

List of recommendations and planet visualization

User testing at RBB and LMU

The user research for this prototype was conducted in the offices of RBB and LMU in Germany. It included professional users (19 participants) as well as end users (10 participants).

  • The lab test with professional users was intended to address general questions about recommendation systems, as well as assess the recommendation prototype.

  • The user test with end users was intended to provide an overview of how general recommendations are perceived. The testers completed a questionnaire on the general recommendations of audio content.

USer test at rbb

USer test at rbb

Results of testing with professional users

In the questionnaire, 40% of the professional users indicated that recommendations for alternative stations are interesting but only 10% would like to receive those on a daily basis. For these professionals, the most important station and programme characteristics that need to be considered in recommendation engines are:

  • music genres (80% agreed),

  • frequency distribution of specific audio such as: 50% reportages, 30% music and 20% sports (60% agreed)

  • music epoch (55% agreed)

For 24% of the participants, listening habits and history should be considered as well.

For most participants, the title, logo, current playlist and song are most important in selecting a recommended station. Detailed EPG information is not of such great importance.

Location and timing when using the recommender

Location and timing when using the recommender

When evaluating the prototype, participants indicated that they liked:

  • The displayed service info when hovering over a specific service.

  • The displayed service logos and information.

  • The sorting of recommended services.

70% of the participants preferred the list as a visualization of the result, whereas 20% preferred the planet visualization. 10% would like to use a completely different visualization. The participants mainly criticized the following aspects of the planet visualization:

  • When hovering over a service item in the list, the corresponding circle in the planet visualization is not highlighted.

  • The circle radius should correspond to the degree of similarity and not to the length of the name.

  • The visualization is rather imprecise and does not really serve a purpose.

Results of testing with end users

The testers only had slight reservations against recommendations. They indicated that they were concerned about getting the same recommendations all the time. When asked what type of content should be recommended, the testers indicated the following:

  • Traffic (80%)

  • News (60%)

  • Tips for events (60%)

  • Cultural events (40%)

In addition, in considering radio station characteristics to make a recommendation, the majority of testers were in favour of music genres such as pop, rock (80%) or music epochs such as the 80s and 90s (60%).

What’s next?

Working towards the second pilot phase, we aim to integrate features into 2 HRADIO prototypes: a car app and a mobile app. The features that were tested in pilot 1 will be integrated, but also new features such as music substitution will appear. In terms of user involvement, the testing will take place in a semi-controlled rather than a lab environment and we aim to engage at least 100 users for the second pilot phase.

Interested in integrating the recommendations feature or other hybrid features? Get in touch or subscribe to our mailing!