Return-to-office mandates are sold as a fix for culture and performance. The numbers tell a less convenient story. Two of the cleanest experiments we have point the other way: give people partial control over where they work and quits fall, without a hit to output. The unresolved question is how to operate that at scale without slipping back into presence-based management. The answer is not more meetings. It is asynchronous management built on documentation and clear measures of work.
What rigorous tests actually found
In a randomized controlled trial inside a Chinese travel company’s call centers, employees who volunteered were randomly assigned to work full-time from home or remain in the office for nine months. Remote workers performed 13 percent better, mostly because they worked more minutes per shift and benefited from a quieter environment, and their attrition fell sharply. There was a cost. Conditional on the same performance, their promotion odds dropped during the experiment. When the firm later let everyone choose their setup, the gains roughly doubled to 22 percent, a mix of better selection and learning (Bloom, Liang, Roberts and Ying, 2015). The lesson is not that remote is magic. It is that choice and measurement matter.
A separate pre-registered randomized trial run in 2021–2022 among more than 1,600 university-trained professionals at Trip.com tested a model that now sits at the center of corporate fights: two days a week at home, three in the office. The result was a one-third decline in voluntary quits, from 7.2 percent in the control group to 4.8 percent in the hybrid group. Performance reviews over two years were statistically indistinguishable between groups, and promotion rates showed no difference either (Bloom, Han and Liang, 2024). The design caveat is important. This was one firm, with educated workers, in China. As evidence of causation in a real workplace, however, it is unusually strong.
Managers looking for the business case will find it stated in plain financial terms by the authors themselves. In their managerial summary, they report that the reduction in turnover saved Trip.com millions of dollars in recruiting and training costs and recommend that firms A/B test their own practices rather than argue in the abstract (Bloom, Han and Liang, 2024). The fact pattern is consistent across the two studies. Fully remote raised measured output but carried a visibility penalty for advancement during the trial. Hybrid cut attrition with no measured loss to performance or promotion. The common thread is control over where work happens, paired with credible measurement of what gets done.
Async is an operating system, not a schedule
This is where most return-to-office debates falter. They fixate on location and ignore how work moves when colleagues are not co-located or online at the same time. Asynchronous work is the mechanism that keeps distributed teams productive. GitLab, which has operated one of the largest all-remote workforces, states the core dependency bluntly in its public handbook: asynchronous collaboration is unworkable without strong documentation because, absent a written record, a single follow-up question can stall a colleague for hours or days. The company treats a living handbook running to thousands of pages as the operational backbone that lets people self-serve answers instead of waiting for someone to come online (GitLab handbook, 2026). This is self-reported practice, not a randomized trial. It is, however, a coherent account of the infrastructure a distributed system requires.
The experiments do not test asynchronous methods directly, but their implications are clear. If people spend less time in the same place at the same time, the work product must be observable without a manager in the room. That pushes organizations to shift from measuring attendance to defining outputs, documenting decisions and processes so that information flows without meetings, and sequencing handoffs so time zones become a feature rather than a tax. Without that operating discipline, hybrid quietly reverts to synchronous habits, and the promised retention gains wither as days at home fill with video calls and status checks.
The implementation discipline
Most firms do not need another manifesto. They need guardrails. The authors of the Trip.com hybrid experiment spell out three conditions for success that generalize well beyond one company: rigorous performance management, coordinated team schedules for in-person days, and visible top-leadership support (Bloom, Han and Liang, 2024). Each is a practical design choice.
Rigorous performance management starts with defining units of work that can be assessed without proximity. Outputs, service levels, and cycle times beat impressions. Coordinated schedules matter because hybrid is not a free-for-all. Teams need overlap for onboarding, problem solving, and social capital. Leadership support signals that managers will be held to account for outcomes, not attendance, and gives air cover to retire legacy rules that conflate face time with contribution.
The 2015 call-center trial carries a second lesson: visibility and promotion. In that experiment, remote workers were less likely to be promoted, conditional on the same measured performance, during the trial window (Bloom, Liang, Roberts and Ying, 2015). That finding argues for explicit calibration. If leaders want hybrid or distributed work to endure, they must audit promotion criteria, demand written evidence for advancement decisions, and ensure that sponsorship does not default to those physically present. When the same firm later let employees choose their setup, productivity gains rose, suggesting that allowing fit and iterating on practice matters as much as the policy headline (Bloom, Liang, Roberts and Ying, 2015).
What leaders should do next
Stop treating hybrid as a perk and start treating it as an audit of management capability. The best causal evidence we have says two days at home can reduce quits by roughly a third without denting performance or promotion in a professional setting, with millions in likely savings from avoided turnover according to the company studied (Bloom, Han and Liang, 2024). The field trial a decade earlier shows that fully remote can raise output while imposing a career cost if visibility is unmanaged, and that choice amplifies gains when the system improves (Bloom, Liang, Roberts and Ying, 2015). To make these benefits stick, build the asynchronous backbone: document how work is done, publish decisions, and design handoffs so people can move work forward without a meeting. Practitioners who run this way insist the documentation is the point, because it lets trust scale when people are not co-located (GitLab handbook, 2026).
A simple starting plan fits on one page. Pick two pilot teams. Define three measurable outputs each. Coordinate two in-office days for the pilots. Establish a light but real review cadence on delivery. Move recurring updates to written memos. Invest modestly in a shared handbook and hold managers to updating it weekly. After one or two performance cycles, compare attrition, onboarding speed, and delivery against a matched control, and publish the results internally. That is how you turn a contested policy into an operating advantage.
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 from 7.2% to 4.8% with no loss to performance or promotion, grounding the case for hybrid. https://www.nature.com/articles/s41586-024-07500-2
- Bloom, Liang, Roberts and Ying (2015). Does Working from Home Work? Evidence from a Chinese Experiment, Quarterly Journal of Economics. A randomized trial at Ctrip where full-time WFH raised performance 13% and halved attrition but reduced promotion odds during the trial, highlighting the visibility risk and the value of choice when gains later doubled to ~22%. https://academic.oup.com/qje/article-abstract/130/1/165/2337855
- Bloom, Han and Liang (2024). One Company A/B Tested Hybrid Work. Here's What They Found, Harvard Business Review. The managerial write-up of the Trip.com RCT that reports estimated millions in turnover savings and sets out three conditions for making hybrid work. https://hbr.org/2024/10/one-company-a-b-tested-hybrid-work-heres-what-they-found
- GitLab (2026). Handbook-first & asynchronous work, GitLab public handbook. A practitioner case asserting that async is infeasible without strong documentation, supporting the operational shift from presence to written, self-serve knowledge. https://handbook.gitlab.com/handbook/company/culture/all-remote/asynchronous/