The BBC wants to bring personal video recommendations to its iPlayer catch-up service, and it’s been testing a number of methods and algorithms as part of a multiyear research project. The broadcaster just finished a final field test, and its lessons learned are worth a look.

bbc mymedia trial

How can TV viewers discover new content online without relying on traditional programming guides? That’s the question at the core of a recent BBC research project aimed at improving the experience of its iPlayer catchup service.

The iPlayer service has proven to be hugely successful, racking up close to 1.4 billion media requests in the first 11 months of 2010 alone. However, the service doesn’t currently use any personalized recommendation techniques, and the BBC wanted to find out how the introduction of recommendations would impact the iPlayer experience and media consumption in general.

The broadcaster has been participating in a pan-European multiyear research project dubbed MyMedia, and it recently concluded a iPlayer personalization field trial that involved testing various approaches to personalization with a panel of 60 users. The trial involved presenting users recommendations based both on their mere usage of the iPlayer (think Last.fm) as well as explicit ratings (think Netflix). It also tested broad recommendations as well as very specific tips based on genres or channels, and published a report about its findings this week (PDF).

Some of the lessons learned were predictable, but others were surprising:

  • People love to give direct feedback. Developers sometimes argue that implicit feedback based on usage behavior is better suited for content recommendations because it requires less work on part of the end user, but the participants of the BBC trial actually were more satisfied when they could actively rate items.
  • Older people vote more. Much like with U.S. elections, voting on content seems to be more popular with older than with younger users.
  • More is better. Users preferred to have a higher variety of recommendations, which the BBC provided by offering specific tips based on genres, type of content etc. Some users said that they liked to see familiar content within the recommendations, but in general, people preferred to be surprised.
  • Sharing is good. The BBC didn’t allow users to share their recommendations with their peers, but most would have been up for it. People weren’t really all that concerned about potential privacy implications.
  • Recommendations are essential. That’s probably the biggest lesson, even though it may seem like a no-brainer: All of the users participating in the BBC’s field test thought recommendations were “very useful,” and many said it took the pain out of finding videos to watch.

Companies like Netflix have spent years perfecting their recommendation algorithms, and many are looking to integrate Facebook and other social networks to provide even better recommendations.

The BBC field test shows that these efforts are worth it, but it also seems to suggest an even more basic lesson: Any type of personalized recommendation is better than providing no recommendations at all — and getting it right doesn’t actually seem all that hard. In the test, participants reported improvements in content discovery after as little as three days of rating content.

Related content on GigaOM Pro: (subscription required)

  1. BBC iPlayer has had personalised recommendations for ages.

  2. iPlayer recommendation lessons | The Filter: Team Blog Friday, February 4, 2011

    [...] Roettgers’ excellent article on video recommendations this week highlighted the findings of the BBC’s recent MyMedia project research. Some of the [...]

  3. The recommendation finding makes sense to me – I think that the real change will come when TV becomes more social in an unobtrusive way – people take much more heed of recommendations from people they know. I’d be curious to see whether the iPlayer let to more tv watching overall.

  4. [...] some of that mojo with its own personalized recommendations system. But to do so, it will need to get recommendations right — otherwise it risks alienating viewers rather than getting them to watch [...]

  5. [...] Recent research from the BBC has shown that content recommendations have a significant impact on the time people spend watching video online, but companies like Netflix have in the past struggled to add social elements to these recommendations. Hunch’s approach of using known patterns to predict unknown preferences could help to make these types of recommendations much more valuable. [...]


Comments have been disabled for this post