Testing the Path to a Redesigned Comments Feature
Figma has rolled out a series of updates to its comments feature over the past two months, and the changes were driven as much by data science as by design. Emily Jia, a data scientist at Figma, outlined how a series of experiments—some of which defied initial assumptions—shaped the final product.
The work began when a cross-functional working group focused on the company’s Starter plan started looking for experiences that unlock collaboration. Multiplayer editing is the most visible form of real-time teamwork in Figma, but the group hypothesized that comments would be the real driver of engagement for a broader team. That hypothesis was partially validated: data showed comments are a strong indicator of team growth and engagement. Yet the same data revealed comments weren’t widely used, which prompted a deeper investigation.
The First Surprise: It’s a Discoverability Problem
User research sessions pointed to a simple explanation: users found comments valuable but couldn’t locate the feature. At the time, users had to enter comments mode by clicking an icon in the upper left corner of the editor, an obscure location for many.
To confirm this, the team ran an experiment targeting developers with view-only access to files—a group that depends on close collaboration with designers but historically avoided comments. After two weeks with a 50/50 split, comment creation in the test group jumped 45% when users were prompted to try the feature. The increase didn’t cannibalize future usage; users who discovered comments were just as likely to return the following week.
A Reverse Result Shifts Strategy
The second experiment tested a new home for the comments entry point. The hypothesis: moving it from the left side of the editor to the right side, where collaboration features live, would make it more prominent for cross-functional partners. The result was a 20% drop in comments discoverability across new users within seven days of signup. The change was small but the effect was significant.
The outcome was a reminder that most product hypotheses are wrong. Microsoft’s experiment platform team found in 2009 that only about a third of ideas improve the metrics they target, and other large tech companies report similar success rates. Rather than continuing to chase a fix for discoverability alone, the team broadened its approach to improving how users manage, sort, and interact with comments once they find them.
Rolling Out a New Launch Paradigm
As the product design iteration continued, the data, product, and go-to-market teams decided to apply the same experimental rigor to the launch itself. The result was an incremental rollout over several weeks, timed with a public announcement at the outset to set user expectations.
Randomization wasn’t straightforward. Since comments are collaborative, the team randomized entire teams on the Professional and Organization plans rather than individual users—having two teammates see different versions of the same feature would distort the experience. That setup required a larger test group to maintain statistical power because usage is correlated within teams.
What the Data Showed After Launch
The updated comments experience produced significant gains. Across all users, the rate of leaving comments rose, with the largest increase among non-designers. Internally, Figma employees saw the same trend, tripling their comment activity since early 2021.
The data also pointed to quicker, more responsive collaboration. While the median comment response time remains 1-2 hours, the 75th percentile response times have dropped by several hours. That matters for Figma’s goal of making the design process approachable for anyone on a product team, not just designers. By making comments more visible whenever a file opens, the feature now encourages a broader range of voices to participate.
For the data science team, the project became a playbook for future launches. The approach extended from setting metrics with leadership to advising product and design teams—and the results validate that consistent experimentation can guide a product from experiment to launch.
