This episode is in English. Mike Rooseboom is the Product Manager for one of the Growth teams at Epidemic Sound. At Epidemic Sound content creators can find royalty-free music for video or sound productions such as Youtube videos or podcasts.
Mike’s Growth Team consists of him as the Product Manager, a Data Scientist, a Product Designer, a Tech Lead, several Full-Stack Engineers, a Machine Learning Engineer, and an Engineering Manager. The team works very closely with data scientists and user researchers and tries to work systematically with data and insights, both quantitative and qualitative, to understand user behavior and where to optimize. From insights they derive hypotheses and bets, from which they create experiments or activities to validate if they’re on the right track or not. Usually, their teams have data scientists embedded, thus part of the day-to-day. User research is something they get help from centrally, but also do a bit themselves in the team, with QA help from the central team.
Based on both quantitative and qualitative insights, they believe visitors and users don’t really understand what it is that they provide. It seems simple enough, but music licensing is hard and it can easily be mistaken for music streaming. Their hypothesis was that visitors needed a bit more information on how this benefits them. So they did different things. Add a page about licensing information. Add a quiz on what subscription works for you. Add information per user segment, addressing their benefits and pain points directly (Youtuber, Streamer, etc). Apart from licensing, which is still hard, it resulted in more signups but also more retention and lower customer service cases. What they learned was that if you address your visitors according to their needs and pain points, you build trust and motivation, even if it seems like you’re adding friction to the funnel.
A challenge they had was that the website was quite slow, especially the further away you are from their servers in Europe. Google looks at The Core Web Vitals – Largest Contentful Paint (LPC), First Input Delay (FID), and Cumulative Layout Shift (CLS) for ranking. Epidemic Sound looked at the correlation between LCP, and conversion rate. The result is a much faster site and improved Google’s page speed index score. And they learned that it is hard to attribute the work to ranking or signups, or even any qualitative input from users. But looking at their own data, they know the site is faster around the world than it used to be.
One insight is that they keep hearing that people want to hear the music before signing up, and they have historically kept the music a little hidden, like Netflix or Viaplay, but this seems to annoy people. So they made the music available through a more robust top navigation. This would also improve SEO. And the results were…Terrible, it killed conversion. They did this many times in different iterations, and they are still not done. Experimentation isn’t about winning, it’s about learning. They can continue to iterate because they believe that there is something here.
Datadrivet is a podcast by scilla.studio. Do you want help getting started with experiments in your team? Get in touch with us at scilla.studio or on Linkedin. The podcast hosts are Joni Lindgren and Jasmin Yaya.