Spotify expands its discover algorithm to podcasts

Spotify has signalled its intention to allow its discover algorithm to help personalized podcast recommendations. This new innovation comes with the launch of its Your Daily Podcast playlist, which will offer regular Spotify users an automated list of podcasts on a daily basis.
Desktop version of discover weekly on Spotify
Spotify’s Discover algorithm has been a massive success and the company now hopes it can do the same for Podcasts. Photo: Spotify

New playlists

The new playlist concept has been tested behind the scenes, and Spotify has been positive about the results achieved. With this in mind, the company is now ready to roll out the feature, with users in the US, UK, Germany, Sweden, Mexico, Brazil, Canada, Australia, and New Zealand benefiting first.

Spotify’s playlist mixes podcasts that people follow with the recommendations of the algorithm itself. The full details of how the algorithm calculates recommendations have not been revealed, but it is known that the system primarily considers listening history in order to make suggestions.

According to the Director of Product Design at Spotify, Emily Rawitsch, the playlists will feature a smorgasbord of different content, with longer podcast episodes included, alongside featured trailers and small morsels of news. Spotify has announced that it will not commit to a particular format in the foreseeable future, with no preferential treatment included for certain shows or networks.

Spotify podcasts
The new algorithm based feature on Spotify has been named ‘Your Daily Podcasts’. Photo: Spotify.

Variety of choice

Rawitsch also confirmed that listeners may be shown an episode in a series that they’re already subscribed to, a similar topic to something they already enjoyed, a popular show in a genre they have used previously, or a program featuring people for whom they’ve displayed a preference.

Spotify is keeping its cards pretty close to its chest with regard to granular insight, or how the new algorithm will determine what content will be promoted. While the system already intelligently sources music, podcasts are by their very nature more difficult for algorithms to understand and classify, so extensive testing must have taken place already.

It is possible that a team working for Spotify has transcribed shows in order to assist the algorithm with understanding. But the company has been reluctant to confirm any such suggestion. A spokeswoman instead suggested that Spotify will focus on listening behaviour and history, with the goal of expanding further as recommendations become more prominent.

Feedback loop

Once the system goes live, users will be able to offer their feedback to Spotify, and this will help the music and podcast-driven site to refine the process. It is reasonable to expect some teething troubles in the early days, but one of the advantages of machine learning is that the artificial intelligence involved rapidly refines its approach.

Spotify mobile
Music streaming has become big business in recent years, and Spotify’s annual revenue has already grown to $5 billion. Photo: Spotify.

One of the challenges for Spotify will be delivering playlists that genuinely appeal to users. The way that podcasts are consumed is far less predictable than music, and some podcasts last for several hours; making the refreshing of a playlist to be complex logistically.

But content makers may welcome the opportunity for their work to be pushed towards accepting ears, and in what is a hugely competitive marketplace this will help liberalize the environment to some extent.

0 0 vote
Article Rating
Notify of
Inline Feedbacks
View all comments
You May Also Like