Developers held steady through the COVID-19 shift

When the COVID-19 pandemic forced millions of people to stay home, software development teams—like most organizations—had to adapt to remote work almost overnight. GitHub’s data science team analyzed developer activity on the platform in the early months of 2020 to understand how this abrupt transition affected productivity, work habits, and collaboration.

The analysis compares the first three months of 2020 with the same period in 2019, focusing on three areas: productivity and activity, work cadence, and collaboration. The findings offer a snapshot of how developers responded to disruption, with implications for team leaders navigating distributed work.

Key findings

  • Developer activity held steady or increased year over year. Per-user metrics including pushes, pull requests, reviewed pull requests, and commented issues all showed slight increases compared to 2019, suggesting developers maintained their output despite uncertainty.
  • Issue activity fluctuated with virus outbreaks and shelter-in-place orders. In enterprise repositories, GitHub issues rose and fell in correlation with local COVID-19 responses, likely reflecting the disruption of normal coordination routines as teams moved to distributed work.
  • Work days got longer. Developers’ active hours extended by up to one hour per day, on both weekdays and weekends, possibly due to interruptions from family or childcare responsibilities at home.
  • Extended activity raises burnout concerns. Sustained engagement may be occurring at the expense of personal time and recovery breaks. The report cautions that this tradeoff may not be sustainable over the long term.
  • Collaboration increased, especially on open source. Several open source projects saw activity spikes, and the time to merge a pull request dropped compared to the prior year—a sign that developers were spending more time working together on community projects.

Implications for teams and leaders

The data suggests developers are resilient: productivity remained consistent or rose through the crisis. For organizations, that resilience can be reinforced by supporting flexible ways of working and keeping tools in place for planning and tracking enhancements, tasks, features, and bugs.

Open source also emerged as an outlet for learning, creativity, and community engagement during lockdowns. The report recommends that companies review policies around external work and technology so employees can participate in learning programs or outside projects without friction.

With longer work days and added home responsibilities, burnout risk is real. The report’s advice to individuals and managers alike: take breaks, build a sustainable workload, and maintain clear boundaries between work and personal time.

GitHub plans to revisit these themes in its annual State of the Octoverse report later in 2020, which will show whether the trends observed in the early pandemic persisted over time.

Developer activity: a composite view

Developer productivity cannot be reduced to a single number such as lines of code or issues closed—those metrics ignore the impact, complexity, and outcome of the work behind them. Research on developer productivity points to a combination of factors: time spent writing code, completing code, and reviewing code. We follow that approach by tracking several forms of GitHub activity per user: pull requests, pushes, reviewed pull requests, commented issues, and issues created.

Development doesn’t change

Development activity on GitHub is a good proxy for activity that is fairly robust to shifting work routines. While some things like kanban boards may have moved from whiteboards to online tools, development can be observed via the same pushes, pull requests, and issues regardless of whether we do that work from an office or at home. This means our year-over-year comparisons are reliable. In contrast, the use of video conferencing increases when people can’t meet in person, even if the number of meetings stays the same. This makes year-over-year comparisons unreliable for people who were not previously working from home.

Theme #1: Data

The data for this section of the report comes from analyzing all GitHub platform activity—public (including open source) and private activity—year over year. The period of comparison is January 1, 2019 – March 31, 2019 vs. January 1, 2020 – March 31, 2020 respectively.  The change in the geographic distribution of active users included in the analysis year over year is shown below. Figure: Geographies included in Theme #1 analysis: Distribution of active users January 2019 – March 2020
Update—May 6, 2:00 pm PT: This chart was replaced to reflect the corrected percentage for Australia.

Pull requests, pushes, reviews, comments: steady or up

Comparing activity year over year, the picture is consistent for all four of those activity measures: activity per active user is equal to or higher than in 2019 across the first quarter of 2020. Pull request creation per user shows the expected dips on weekends, but after removing weekends for readability, the increase is clearer.

Push volume tells the same story: higher than last year and sustained through the first three months of 2020. Reviewed pull requests and commented issues follow a similar pattern. The initial reaction to this data suggests developers kept up their output, or even increased it, during the early pandemic shock.

Issues: a more complicated story

Issue creation per user did not follow the same trajectory. For most of 2020 it was lower than 2019, dipping further in February, before recovering in mid-March. Breaking the numbers down by repository ownership type reveals different behaviors: repositories owned by free user and paid team accounts saw the largest increase in issue creation rates, while Enterprise Cloud-owned repositories followed a different path.

For enterprise repositories, that path had two notable phases. Through mid-February, issue creation dropped noticeably—corresponding to when Asia and Europe were hard hit and the US West Coast began moving to remote work. Late February and early March saw activity recover to 2019 levels, though weekends remained quiet in a way they did not for the other repository types. In the second phase, issue volume climbed past last year's levels, a likely sign that teams resumed the planning and coordination cycles of enterprise software development.

The swings make sense if you consider what issues are for: issues are the record of communication, planning, and task tracking that structure much of development work. In personal or hobby projects, that planning often never gets written down—a sticky note may be enough. Enterprise development depends on structured, coordinated planning across teams, which means issues are directly sensitive to disruptions in the working patterns that surround them.

Our reading of the sequence is:

  • Teams scrambled to adapt to working from home; meetings that feed the issue backlog slowed or stopped—issues kept ticking over.
  • Developers worked through the existing backlog of issues as they adjusted—issues slowly declined.
  • Teams rebuilt their ceremonies—standups, kanban boards, ad-hoc discussions—in distributed form; issues stabilized or started growing again.
  • With planning re-established or the backlog nearly exhausted, teams began re-logging upcoming work into issues—issues spiked.

The recovery from the February trough was relatively quick. Given the scale of the disruption and ongoing uncertainty, though, we expect further fluctuations before activity settles into a new equilibrium.

Work cadence: longer days, more pushes

Activity levels only tell part of the story. To see how the shape of a developer's day changed, we defined a "work day" as the time between the first and last git push to the default branch. That is a rough proxy, but it gives a way to see when work starts and ends day over day.

Theme #2: Data

The data for this section of the report comes from analyzing paid organization accounts that meet the following criteria:
  • Created before January 1, 2019
  • Had activity each month through March 2020
  • On a paid Team or GHEC account
More than 40,000 organizations were included in our analysis, with strongest representation in North America (40%), Europe (35%), and Asia (17%) as shown in the following figure: Figure: Geographies included in Theme #2 analysis: Distribution of active paid accounts, January 2019 – March 2020

We focused the analysis on the US Pacific and Eastern time zones. Both coasts issued coordinated shelter-in-place orders during the period, and there were enough developers in each zone to produce meaningful results. We looked at both the length of the work day and the volume of pushes within it, so we could distinguish coworkers who stretch a day out from those who are doing more work.

Pacific time zone: longer days, heavier volume

Developers in the Pacific time zone had variable day lengths through the start of 2020, then from mid-March the work day lengthened noticeably—typically by 30 to 60 minutes. Much of the growth happened on weekends.

The extra time could mean more actual work rather than a stretched day of small bursts. Push volume per day also grew year over year, especially in March. For developers used to commuting, that time—officially reclaimed, unofficially up for grabs—can go to either home duties or work. Higher push counts point to contributors doing more regular, sustained work. That could reflect understandable motivations: concern about job security in an economic downturn, using work as a steadying distraction while stuck at home, management pressure to hit release targets, or a wish to keep delivery cadences predictable when everything else is not.

Eastern time zone: shorter baseline, weekend growth

On the East Coast, work days were shorter than last year for much of 2020, before turning upward in mid-March like the West Coast, though to a smaller degree—an increase of 15 to 30 minutes, again largely concentrated on weekends.

Push volume for East Coast developers showed a different curve: higher than 2019 through February, before declining. During the shelter-in-place ramps, East Coast developers shifted their distribution of work—more pushes spread through the day and week—without the sustained volume rise seen in the Pacific zone.

Burnout

Our analysis suggests that developers are continuing to do sustained and even increased amounts of development, which some may cheer as evidence that productivity has continued in the face of uncertainty. However, combined with our work cadence analysis, we caution that developers, leaders, and organizations should take proactive steps to prevent burnout, and watch for it among their teams and peers. The World Health Organization has recognized burnout as “an occupational phenomenon resulting from chronic workplace stress that has not been successfully managed.” While burnout specifically refers to workplace stress, it can be difficult to manage right now when our work is invading our personal space. Dealing with burnout is important for our mental well-being, both in the workplace and in our personal lives. Teams and leaders that support flexible and sustainable work schedules and watch for burnout will have colleagues and teams that are happier and more productive. Remember, we’re all in this together. For more on burnout and what you can do to address it, we point you to the article Understanding Job Burnout from Dr. Christina Maslach, an expert on workplace burnout. If you prefer, you can watch or listen to her conference talk from DevOps Enterprise Summit 2018 in Las Vegas.

Collaboration in a Remote-First World

Software development is rarely a solo act, even if popular culture often frames it that way. Remote work has forced teams to replace in-person whiteboards and sticky notes with digital equivalents, and the rhythms of communication and coordination have shifted accordingly. This final section looks at how those shifts show up in collaboration data.

Pull Request Merge Times Shift Earlier in the Year

Pull requests are the primary mechanism developers use to propose changes, and the time it takes to merge one captures the full collaborative loop: review, discussion, follow-up commits, and ultimately integration. In January 2020, merge times were notably slower than the year before. For repositories owned by Enterprise Cloud accounts, pull requests took four to five hours longer to merge than in January 2019, while paid Team accounts saw roughly one hour longer merge times. There's no clear explanation for that early-year slowdown.

By March, the picture flipped. Enterprise Cloud repositories saw merge times that were only 15–30 minutes longer year over year—an improvement of three hours or more from January. Paid Team accounts returned to about the same merge times as the previous year, a roughly one-hour improvement from January. That drop, coinciding with widespread shelter-in-place orders, suggests developers were more available to review requests, had more focused time, and could respond faster while working from home.

Open Source Activity Picks Up

Open source repositories showed a similar pattern. Merge times early in 2020 were a few hours slower than in 2019, but by March they were faster by anywhere from 45 minutes to nearly four hours. That responsiveness points to deeper engagement with open source projects as developers looked for work they could tackle from home.

The engagement went beyond faster reviews. Open source project creation grew throughout the period, with 27.62% more new open source repositories created in late March 2020 compared to the same time last year. Collaboration on top public projects, measured by distinct contributors, also jumped in mid-March—the same inflection point seen in pull request merge times.

The community response to public COVID-19 projects was particularly notable. Contributions began ramping up in January, with another surge in contributors around mid-March.

About the Data

Activity on GitHub grew compared to 2019, and that increase was steady through the first quarter even as the pandemic unfolded. The growth is in line with patterns from prior years, which is notable given the broader economic uncertainty. Exact user counts are withheld for confidentiality, but the visual trends show year-over-year comparisons with regular dips on weekends.

The analysis covers developer activity from January through March 2020, matched to the same period in 2019. To ensure comparability, dates were aligned by day of the week, shifting 2019 data by two days to account for yearly drift and leap year. The three-month window was chosen to focus on when COVID-19 first disrupted work routines, avoiding seasonal variability that would come with a longer comparison. All data is reported in aggregate, with sample details provided per section but no raw data released to protect anonymity.

Acknowledgements

The authors thank the data scientists, reviewers, copyeditors, and designers who contributed to this work:

  • Data Scientists: Greg Ceccarelli, Anna Filippova, Taylor Holland, Derek Jedamski, Scot Kelly, Rowan Wing
  • Reviewers: Martin Fowler, Sam Guckenheimer, Caitie McCaffrey, Rachel Potvin
  • Copyeditors: Mandy Campbell, Cheryl Coupe
  • Designers: Jobey Greenwood, Sabrina Majeed