Product-led growth at Epidemic Sound, with Mike Rooseboom
In this episode of Datadrivet, Joni Lindgren and Jasmin Yaya talk with Mike Rooseboom, a product manager on one of the growth teams at Epidemic Sound, about how growth works when it lives inside product.
At Epidemic Sound, growth sits in the product function rather than marketing, and each team has an embedded data scientist. The teams combine quantitative and qualitative signals to find where users get stuck and where there is room to improve.
A core finding was that visitors struggled to understand the service. Many confused music licensing with streaming. The team responded with licensing information pages, a subscription quiz, and benefit messaging tailored to different segments. Those changes improved signups and retention and cut down on customer service cases. Separately, work on website performance through Core Web Vitals (LCP, FID, CLS) correlated with better conversion.
Not everything worked. An experiment to make music more discoverable through navigation killed conversion, and it kept doing so across several iterations. Mike’s framing is that experiments are a way to learn, so a result like that still tells you something worth knowing. The hosts and Mike discuss product-led growth as typically built on a freemium model that shows the product’s value, while noting that some businesses still need sales or marketing, and that product and marketing collaborate closely on growth.
The takeaway is that growth inside product is steady work: find the confusion, fix the friction, measure the result, and treat the experiments that fail as part of the learning.
Listen to the full episode of Datadrivet for the full conversation with Mike Rooseboom. If you want to see how your own growth and retention compare to others in your model, that is what the benchmark tool is for: https://benchmark.scilla.studio
See where your numbers actually land
Plot your retention, CAC payback, LTV:CAC and K-factor against the B2B and Consumer bands, and find out whether a good-looking number is real or sitting on a leaky retention curve.
Run the free diagnosis →