The Big Holes in GitHub’s Remote Work Data
In May, GitHub released a report that found that “developer activity remains largely consistent or increased compared to last year” as developers have “continued to contribute and show resilience in the face of uncertainty.”
New data from Software’s community of 100k developers shows the GitHub data is incomplete.
Many developers are getting left behind during the shift to remote work.
According to our data, developers who code well at the office are more likely to code well at home.
That may not apply to developers who typically code less. Lower output developers see significant drops in coding when switching from office to remote work.
GitHub’s activity data falls short
GitHub’s remote work data only tracks development activity beginning at each commit.
Everything that happens before commits—like code time in the local branch—is lost forever.
Commit activity is also subject to skew. Smaller and easier changes pushed to GitHub will show increased “activity,” even if coding and feature output falls.
GitHub’s data also does not break out its community into cohorts.
Lumping together all developers hides the true impact of remote work on different types of developers.
Pre-commit data shows the impact of remote work is uneven
Data from the Software community shows that if you code a lot at the office, you likely code a lot at home.
When switching from office to remote work, highly prolific developers—those who are in the top half in terms of coding output—produce 6% more code.
They also work the same number of hours at the office and at home.
Active code time per developer—defined as the time spent actively writing and editing code in your editor or IDE—holds steady at 2.6 hours per day when switching from office to remote work.
Developers who don’t code a lot at the office, code even less at home. Developers who code less at the office are hit harder when switching to remote work, with more than a 15% decrease in coding output.
They also code fewer hours. When switching to remote work, active code time falls nearly 22%, from 1.4 hours to 1.1 hours per day.
Deciding the future of work with data
Tech CEOs are taking polarized views on the future of work. Consensus is hard to find. Some ended their office leases, while others doubled down on plans to reopen their campuses.
GitLab and HashiCorp shun office culture in favor of a remote-only workplace. Twitter, Square, and Shopify are embracing remote-first policies.
Microsoft, Facebook, and Google are increasingly remote-friendly, but plan to keep their offices.
Netflix staunchly opposes remote work, its CEO Reed Hastings arguing that remote work is a “pure negative.”
As lockdowns ease and offices reopen, leaders will make the crucial decision of how best to adapt their work cultures. Will your team switch to an all-remote workplace? Will your team mandate that employees commute to offices again? Will your team find a new balance between the two?
With better data, teams can choose a work environment that preserves developer productivity and happiness.
About the results
We aggregated anonymized data from Software’s community of over 100,000 developers to analyze how work location impacts productivity. Data is from developers using our Code Time plugins for VS Code, Atom, Sublime Text, PyCharm, IntelliJ, and WebStorm, among other code editors.