25% more people came back to Skillshare every day after we launched. That number sounds exciting in a deck. The reason behind it is less flashy — and, honestly, more interesting.
We didn’t build a new algorithm. We didn’t redesign the whole app. We asked learners what they actually wanted to see when they opened Skillshare. And they told us something we hadn’t fully admitted to ourselves: they wanted to see what other people were making.
The Marketplace Problem
Skillshare had millions of learners and thousands of classes. By most measures, the content side of the business was strong. But when I looked at how people actually used the platform, something was missing. You’d take a class, maybe share a project, and then… go back to browsing for the next class. The loop was linear. Watch, maybe post, repeat.
That’s a marketplace, not a community. And the difference matters more than it sounds.
Creative learning doesn’t just happen in the relationship between a student and a class. It happens when you see someone else’s watercolor piece and think, “I didn’t know you could do that.” It happens when a beginner’s rough sketch gets three encouraging comments and they come back tomorrow to try again. The social layer is part of the pedagogy — we just hadn’t built it yet.
What the Research Told Us (And What It Didn’t)
We ran structured research using Teresa Torres’ Continuous Discovery Habits framework — weekly interview cadences, opportunity solution trees, hypothesis-led iteration. Not a big research sprint at the start of the project. Ongoing, compounding weekly conversations that kept us honest throughout.
The consistent signal from learners was clear: they wanted to see what peers were creating, not just consume classes. They wanted to follow creators whose work inspired them. They wanted a reason to check in on Skillshare even on weeks when they weren’t actively taking a class.
What the research didn’t immediately tell us — and what we had to work through carefully — was how to serve that need without breaking something important.
“Skillshare’s brand was built on structured learning. Our users came for classes. The risk of adding a social feed wasn’t feature bloat — it was identity drift. We had to make it feel curated, not chaotic.”
That tension shaped every design decision. We weren’t building Twitter for creatives. We were building a feed that felt intentional — one where everything surfaced had some relationship to learning.
What We Built
The Creative Feed became the answer. A personalized, centralized place where learners could discover activity from the people and topics they cared about. The feed surfaces:
- Projects shared by other learners in classes you’ve taken or categories you follow
- Class discussions and teacher updates from creators you follow
- New classes from followed categories and teachers
We also built a full follow system for Skillshare’s 7M+ learners — allowing each person to shape their own creative network. Follow a ceramics teacher whose style you love. Follow the illustration category. Follow a peer whose project knocked you sideways.
The architecture of what gets surfaced when is something we iterated on heavily throughout the rollout.
🚧 Need more context: What signals drove feed ranking and ordering? Was there a personalization algorithm, manual curation logic, or a mix? Any notable decisions about what not to surface?
The Results (And Why They Make Sense)
Daily engagement increased by 25% within months of launch. Which, when you understand the mechanism, is almost obvious in retrospect.
We gave people a reason to open the app that wasn’t “I’m ready to start a new class.” We gave them a way to stay connected to the community between classes — to see what their creative network was making, to get inspired, to stay warm. The learning didn’t change. The belonging did.
What I’d Do Differently
I’d push even earlier to resolve the brand tension explicitly — not just through product design decisions, but through a clear shared statement of what kind of community Skillshare wanted to be. Some of our debates late in the process were really debates about that unanswered question.
I’d also have invested more upfront in understanding the creator side of the follow relationship. We learned a lot about what learners wanted to see. We could have done more work earlier on what creators wanted to share, and how that shaped what felt worth posting.