Return-to-office mandates collide with a stubborn fact. The best experiment we have on hybrid work shows that two days at home reduce voluntary quits by a third with no hit to performance or promotions (Bloom, Han and Liang, 2024). Yet many firms that win this retention dividend lose it back to meeting sprawl and Slack fatigue. The fix is not another day in the office. It is a different unit of management: the written record that lets teams work asynchronously and be judged on what they ship.
What the strongest evidence actually says
In a randomized trial inside Trip.com, 1,612 university-trained employees were assigned by birthday to either five days in the office or a hybrid schedule with two days from home. Over two years of reviews there was no difference in performance ratings or promotion rates. Attrition fell from 7.2 percent in the control group to 4.8 percent in the hybrid group, a one-third reduction that was statistically significant (Bloom, Han and Liang, 2024). This is unusually clean causal evidence for management. It is one firm, a Chinese tech company, and one employee segment, graduates, so generalization should be cautious. But within that boundary the result is clear.
The same authors describe three conditions that made the model work: rigorous performance management, coordinated schedules at the team or company level, and visible support from the top team (Bloom, Han and Liang, 2024, HBR). They also recommend that firms A/B test practices in their own context. Read their list closely. None of the conditions are about furniture or floor plans. They are about legibility: making expectations explicit, aligning time windows, and ensuring leaders reinforce the rules.
The visibility tax of remote, and a plausible remedy
A decade earlier, a different randomized trial at Ctrip, a large Chinese travel agency, found that fully remote call center staff were 13 percent more productive than their office peers, with about nine points coming from more minutes worked per shift and four from a quieter environment. They were also more satisfied and less likely to quit. There was a cost. Remote workers were less likely to be promoted, conditional on the same performance (Bloom, Liang, Roberts and Ying, 2015). When the firm later let employees reselect their arrangement, the measured productivity gains rose to roughly 22 percent, a combination of selection and learning inside that single company (Bloom, Liang, Roberts and Ying, 2015).
That promotion penalty is a warning. If managers do not see how work happens, they default to who is seen. Hybrid reduces quits, but presence bias can still punish those whose best work happens away from the office. The credible antidote is to make output, decisions, and context visible in writing. This is not faith. It follows directly from the conditions specified for successful hybrid. Rigorous performance management requires clear goals, artifacts, and evidence. Coordinated schedules require agreements that are documented and findable. Leadership support requires codified norms that survive turnover.
Practitioners who run large distributed teams have built on this logic. GitLab operates with a handbook-first approach, documenting processes, decisions, and expectations in a public handbook that would exceed 2,000 printed pages. The company states that asynchronous work is impossible without strong documentation, because otherwise a single follow-up blocks a colleague for hours. The mechanism is simple. People self-serve answers from the record instead of waiting for someone to come online (GitLab handbook, 2026). This is not an outcome study. It is a case. But it provides a concrete operating model that aligns with what the experiments require.
Matt Mullenweg, the CEO of Automattic, offers a taxonomy that puts this idea in conceptual terms. At his Level 4, work is asynchronous and judged by output rather than hours or process, a standard many hybrid firms say they want but rarely operationalize (Mullenweg, 2020). He goes further and claims Level 5 will outperform any co-located firm. That last claim is opinion, not evidence. The useful piece is the diagnostic: until output is the unit, you are not truly async.
Make the record the product
If you want the retention gains from hybrid and want to avoid the visibility tax of remote, invest in the record.
- Define what good looks like in artifacts. Tie goals and reviews to documents, code, tickets, and customer outcomes. This is the “rigorous performance management” condition in practice (Bloom, Han and Liang, 2024, HBR).
- Publish how and when your team coordinates. Set the default days in or out at the team level and write down working hours, handoff windows, and the SLA for responses. That is coordinated scheduling, not calendar chaos (Bloom, Han and Liang, 2024, HBR).
- Codify decisions and make them searchable. A decision log beats a dozen meetings. GitLab’s claim is that documentation turns questions into self-service, which is exactly what asynchronous flow needs (GitLab handbook, 2026).
- Train managers to read the record, not the room. The Ctrip trial’s promotion gap, conditional on performance, is a reminder that informal visibility can trump data unless you correct for it deliberately (Bloom, Liang, Roberts and Ying, 2015).
None of this requires expensive software. It requires discipline. Write the meeting agenda and outcomes before you invite people. Refuse status meetings if the status can be read. Time-box real-time collaboration for topics that benefit from live debate, then capture the decision where others can find it. Treat documentation as a core deliverable of leadership, not an afterthought.
The practical read
Trip.com’s attrition drop saved millions in recruiting and training, by the company’s own estimate, and increased profits as a result (Bloom, Han and Liang, 2024, HBR). That is the board-level business case for hybrid. The operating case is that hybrid is fragile without an asynchronous backbone. The experiments tell you where the value is. The handbook shows you how to hold on to it. If your calendars are full but your decision log is empty, you are paying for the office twice. Shift the center of gravity from presence to the record, and let the numbers do the convincing.
Sources
- Bloom, Han and Liang (2024). Hybrid working from home improves retention without damaging performance, Nature. A randomized trial where two days at home cut quits by a third with no loss to performance or promotion. https://www.nature.com/articles/s41586-024-07500-2
- Bloom, Han and Liang (2024). One Company A/B Tested Hybrid Work. Here's What They Found, Harvard Business Review. Managerial translation that names three conditions for successful hybrid and reports the firm's estimate of millions saved from lower turnover. https://hbr.org/2024/10/one-company-a-b-tested-hybrid-work-heres-what-they-found
- Bloom, Liang, Roberts and Ying (2015). Does Working from Home Work? Evidence from a Chinese Experiment, Quarterly Journal of Economics. A randomized trial where full-time WFH raised performance 13 percent and halved attrition but reduced promotion odds conditional on performance; gains rose to about 22 percent when employees later chose their arrangement. https://academic.oup.com/qje/article-abstract/130/1/165/2337855
- GitLab (2026). Handbook-first and asynchronous work, GitLab Handbook. Practitioner case arguing async is impossible without strong documentation and describing the handbook-first mechanism. https://handbook.gitlab.com/handbook/company/culture/all-remote/asynchronous/
- Mullenweg (2020). Five Levels of Autonomy, ma.tt blog. An opinion framework where Level 4 defines async work judged by output rather than hours, a useful diagnostic for hybrid aspirations. https://ma.tt/2020/04/five-levels-of-autonomy/