Return-to-office mandates are a blunt instrument for a subtle problem. The strongest experimental evidence we have says hybrid work improves retention with no hit to output. The real management test now is not place but time: can your teams move work forward without being co-present or constantly online?
What the clean experiments actually say
In 2021 Trip.com randomized more than 1,600 university-trained employees in Shanghai to two days at home per week or five in the office. Over two years, attrition fell by about a third in the hybrid group, from 7.2 percent to 4.8 percent, with no difference in performance ratings or promotions across four semiannual review cycles (Bloom, Han and Liang, 2024). This is a pre-registered randomized controlled trial, the causal gold standard, inside a single firm. The design matters for interpretation. It tested hybrid, not fully remote, and it involved graduate-level roles in a Chinese tech company. Even with those caveats, the retention effect is clear.
A decade earlier, a randomized trial at Ctrip, a Chinese travel company, assigned volunteer call-center staff to work from home or the office for nine months. Home workers were 13 percent more productive, largely because they worked more minutes per shift and benefited from a quieter environment, and their quit rates halved. The penalty showed up in careers: conditional on performance, promotion rates fell for the remote group. When Ctrip later let employees opt in or out firmwide, the measured productivity gains rose to around 22 percent, a combination of selection and learning effects (Bloom, Liang, Roberts and Ying, 2015). The paired findings are the management brief in miniature. Hybrid is a safe bet on retention that does not dent performance. Fully remote can raise output but introduces a visibility tax on advancement if you do not change how you manage.
Hybrid works when management grows up
Trip.com’s executives report that lower turnover saved millions of dollars in recruiting and training costs and lifted profits, a company estimate that explains why boards should care (Bloom, Han and Liang, 2024, HBR). The same authors, writing for managers, are blunt about what made hybrid work: rigorous performance management, coordinated team or company schedules for office days, and visible support from the top team. They recommend firms run A/B tests on their own practices rather than rely on punditry (Bloom, Han and Liang, 2024, HBR). These conditions are not window dressing. They are the scaffolding that turns flexibility into a system instead of a perk.
The Stanford summary of the Trip.com trial underscores who benefited most from hybrid: non-managers saw larger retention gains than managers. It also reiterates the main point executives worry about, namely productivity and promotion. On both counts, the randomized evidence shows no difference between hybrid and office-only groups (Stanford Report, 2024). If your leadership team is still arguing about whether two days at home will tank performance, you are fighting the last war.
Async is the operating system for distributed teams
Place is only half the design. Once teams stop relying on daily co-presence, coordination shifts from meetings to artifacts. GitLab, which runs one of the largest all-remote operations, frames the mechanism plainly: asynchronous work is impossible without strong documentation, because follow-up questions otherwise block work for hours or days. Its public handbook, which would exceed 2,000 printed pages, exists to let any team member self-serve answers rather than wait for a colleague to come online. The claimed benefits include autonomy, efficiency, inclusion, and fewer interruptions, but these are self-reported practices rather than measured outcomes (GitLab handbook, 2026). Even so, the logic matches what the experiments imply. If you cannot manage by attendance, you must manage by clarity of goals, standards, and decisions.
Practitioners have tried to codify what good looks like. Matt Mullenweg of Automattic describes five levels of autonomy in distributed work. Level four centers on asynchronous collaboration evaluated on output rather than hours or process. It is a useful vocabulary for leaders, though it is a taxonomy of practice, not empirical proof, and his strongest claims about performance remain opinion (Mullenweg, 2020). As a diagnostic, it asks a sharp question: do your teams ship to a definition of done that is legible without a meeting?
The visibility trap, and how to avoid it
The Ctrip trial’s promotion penalty is a warning. If managers still infer contribution from what they can see, remote or async workers pay a career tax regardless of output (Bloom, Liang, Roberts and Ying, 2015). The Trip.com hybrid trial did not find any difference in promotions over two years, which suggests that schedule design, shared in-office days, and tighter performance management can neutralize that risk in hybrid settings (Bloom, Han and Liang, 2024). The lesson for distributed leaders is straightforward. Make work visible in artifacts, not in eyeballs. That starts with goals and metrics, continues with written decisions and handover notes, and ends with review processes that read the work rather than the worker’s presence.
The practical implication is uncomfortable and freeing. You do not need to win a culture war over desks. You need to upgrade your operating model so that hybrid is the on-ramp and asynchronous is the default where it fits. Start with a bounded experiment, as Trip.com did, and steal its guardrails: schedule hybrid days at the team level rather than ad hoc, tighten performance management, and have the CEO say out loud that this is the way we work (Bloom, Han and Liang, 2024, HBR). In parallel, run a documentation sprint. Pick one cross-functional process, write the current state down, and require that decisions and next actions live in a shared system. Measure the results you can defend to your board: attrition, time to decision, and cycle time for shipping. The evidence has already told you where value hides. Now build the systems that let you capture it.
Sources
- Bloom, Han and Liang (2024). Hybrid working from home improves retention without damaging performance, Nature. A pre-registered RCT at Trip.com where two days at home cut quits by a third with no loss in performance or promotions. 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 of the Trip.com trial that reports turnover savings in the millions and the three conditions required for hybrid to work. 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 fully remote call-center work raised performance 13 percent, halved attrition, but cut promotion odds; letting employees choose later raised gains to around 22 percent. https://academic.oup.com/qje/article-abstract/130/1/165/2337855
- Stanford Report (2024). Hybrid work is a win-win-win for companies and workers, Stanford News. Secondary coverage of the Trip.com RCT highlighting concentrated retention gains among non-managers and no hit to productivity or promotions. https://news.stanford.edu/stories/2024/06/hybrid-work-is-a-win-win-win-for-companies-workers
- GitLab (2026). GitLab — handbook-first & asynchronous work (all-remote handbook). Self-reported case stating that async is impossible without strong documentation and describing a large public handbook as the foundation. https://handbook.gitlab.com/handbook/company/culture/all-remote/asynchronous/
- Mullenweg (2020). Five Levels of Autonomy, ma.tt. Practitioner taxonomy describing async, output-focused work as a higher level of distributed capability, offered as opinion rather than empirical evidence. https://ma.tt/2020/04/five-levels-of-autonomy/