During the feature’s beta phase, Acast extracted data over a period of two weeks which showed that users are 52% more likely to follow a show if it is recommended to them by the algorithm, and 49% more likely to listen to multiple episodes of a show recommended by Acast.
Johan Billgren, Acast CTO, comments: "Searching for new podcasts is hard and often time-consuming. We want our users to spend their time listening to great podcasts, rather than looking for them, and that is why we are launching Recommendations. Our machine learning algorithm, which we began testing in October last year, gets to know preferences over time by learning from a user’s choices and then recommending new shows that they will like."
The machine learning algorithm analyses anonymised listening data and suggests new content that is relevant to each individual user based on their listening habits and passions.
The algorithm learns over time, meaning that the more a user listens to audio, the better and more personalised the recommendations become; reflecting changing tastes and interests which can develop over time.
As well as users being offered content that is relevant to their listening history, each podcast page displays other thematically similar shows, based on the podcasts enjoyed by other users. As well as enhancing the user experience, Recommendations allows advertisers to target specific demographics with greater precision, by providing them with deeper insights into listening trends.
