Summary

Discover how Skillshare transformed their software development and experimentation process, scaling from slow release cycles to multiple experiments per week using Split’s feature management and experimentation (acquired by Harness).

• Accelerated software releases: Skillshare scaled their release frequency, improving the speed and efficiency of software delivery.
• Enhanced data-driven decisions: With Split’s real-time results and statistical rigor, Skillshare gained insights into feature performance, improving product decisions without hesitation.
• Improved feature rollouts: By using feature flags, Skillshare minimized risk by toggling features on/off and enabling targeted experiments to validate assumptions.
• Streamlined deployment process: Split's seamless integration with Skillshare’s tech stack helped save engineering time, improving deployment safety.

About Skillshare

Skillshare is a vibrant online learning community offering over 30,000 classes in creative fields such as art, design, writing, and photography. With a community of more than 8 million students, Skillshare provides both free and subscription-based courses to help learners enhance their skills. As Skillshare grew, the engineering team faced the challenge of speeding up product releases while ensuring features provided optimal user experiences.

Skillshare's tech stack includes PHP, React, Node, and GraphQL, which support their goal of delivering a smooth experience for students and instructors alike. However, the team was facing bottlenecks in their development process and needed a solution that would allow them to experiment faster and deploy features with minimal risk.

Challenge

Slow release cycles and limited experimentation

Skillshare’s engineering team struggled with slow release cycles—updates being pushed to production only every few months. A key issue was the lack of feature flagging, which would allow for controlled rollouts and safer experimentation. Without feature flags, they deployed large batches of code that were often difficult to test thoroughly before release.

Mike Sherov, Principal Engineer at Skillshare, explains: “Before we had feature flags, we deployed 10,000 lines of code at once. We didn’t have a way to test in production without risking a bad user experience.”

The company was also relying on a rudimentary internal A/B testing system that didn’t provide the statistical rigor needed to make informed, data-driven decisions.

Solution

Accelerated innovation and safer deployments

In 2019, Skillshare adopted Split to enhance their software development processes. By integrating Split’s feature flags and A/B testing, Skillshare could safely roll out new features, run experiments in production, and make more data-driven decisions.

Key benefits with Split:

• Faster experimentation: Split enabled Skillshare to run multiple A/B tests concurrently and gather real-time insights into feature performance.
• Safer rollouts: With feature flags, the team could toggle new features on/off at the press of a button, minimizing risk during deployment.
• Data-driven decision making: Real-time results from Split’s A/B tests allowed Skillshare’s engineers to optimize features based on actual user behavior.
• Dynamic targeting: Skillshare was able to target specific user segments for more personalized experiments and dynamically adjust sample sizes and variations.

Mike Sherov shares: “With Split, we can deploy code more frequently and test in real-time without the risk of exposing all users to potential issues.”

“Being able to test in production has been a game-changer for us. It allows us to learn and iterate faster, which ultimately leads to better products for our users.” — Jean Steiner, VP of Data Science, Skillshare

Results

Scaling experimentation and improving release velocity

The integration of Split led to significant improvements in Skillshare’s software development and experimentation capabilities:

• 12X Increase in release frequency: The team moved from releasing updates every few months to releasing multiple times per week.
• 64+ versions running simultaneously: Skillshare regularly runs 5-6 A/B tests concurrently, with more than 64 versions of the app live in production at any given time.
• Safer feature rollouts: Feature flags allowed the team to minimize deployment risks by testing new features in production without exposing them to all users.
• Improved developer confidence: Feature flags gave the development team psychological safety, allowing them to experiment boldly and deploy with confidence.

Jean Steiner, VP of Data Science at Skillshare, notes: “The ability to test in production, without the fear of breaking anything, has given us the confidence to innovate faster and more effectively.”

Business impact

Faster iteration, better outcomes

By implementing Split, Skillshare achieved faster iteration cycles, better control over feature rollouts, and an ability to base product decisions on real-time, data-backed insights. With the ability to run concurrent experiments and make rapid changes, Skillshare saw measurable improvements in both engineering efficiency and user experience.

Key outcomes include:

• Faster experimentation: Skillshare increased the number of A/B tests, running them more frequently and with higher impact.
• Improved feature quality: By testing in production with feature flags, the team ensured that new features met user needs before full deployment.
• Increased developer efficiency: The team could focus more on innovation, rather than firefighting issues caused by risky deployments.

Future

Building a culture of innovation and experimentation

As Skillshare continues to grow, Split will remain a critical tool for maintaining a high rate of innovation. By using Split’s feature management and experimentation platform, the team can continuously test, measure, and refine product features while ensuring they align with key business goals.

Mike Sherov sums up: “Split has transformed how we approach software development. We can now experiment confidently, release frequently, and make data-driven decisions that help us create a better experience for our users.”