Data-approach as a way to find the best artist for users of musical streaming

Millions of users listen to their favourite artists on the Zvuk app. Currently, onboarding presents users with artists they are least likely to dislike—typically the most popular performers. Recommendations are activated only after users click on the artist. To improve the list displayed and enhance the onboarding experience, we are experimenting with different segmentations, one of which involves segmenting users based on their interests.

The main goal is, firstly, to match users' interests using ecosystem embeddings which reflect the interests. Secondly, to identify clustering applicability and define clusters.

In this talk, I will explain how we approached this, which tools we used, and the challenges we encountered, as well as the conclusions we reached.

The talk will be useful for business and data analysts who work with big data (including data from business partners, etc.).