Return-to-office orders are framed as a rescue mission for mentorship and advancement. Employees hear a different message: face time over output. Both sides are reacting to a real risk. In the cleanest early experiment on work from home, fully remote employees performed better but were less likely to be promoted, conditional on that performance. The newest gold-standard trial of hybrid found no such penalty. The lesson is not that offices fix visibility. It is that well-run hybrid, paired with asynchronous practices that leave a written record of contribution, removes the need to be constantly seen.
What the strongest experiments actually say
A decade ago, a randomized controlled trial inside Ctrip, a large Chinese travel company, assigned volunteer call-center staff to work either at home or in the office for nine months. Home workers were 13 percent more productive, largely from more minutes worked per shift and a quieter environment. They were also happier and their attrition halved. Yet their promotion rate, conditional on the same performance, fell. When the firm later let everyone choose their setup, more than half switched and the measured gains roughly doubled to 22 percent, a mix of selection and learning (Bloom, Liang, Roberts and Ying, 2015). The promotion cost flagged a mechanism leaders still worry about: being out of sight can carry a career price if managers rely on presence as a proxy for performance.
In 2024, a new randomized trial tested the model most firms now debate. Trip.com moved university-trained professionals to a hybrid schedule of two days at home per week, with random assignment by birthday. Over two years, attrition fell by about a third, from 7.2 to 4.8 percent, and there was no difference in performance evaluations across four review cycles. Crucially for the visibility fear, there was no evidence of a difference in promotion rates between hybrid and full-office groups during the same period (Bloom, Han and Liang, 2024). This was not a light-touch pilot. It was a preregistered experiment in a single firm with tight controls. The caveat is important: one tech company in China, a graduate workforce. Still, as causal evidence goes in management, it is rare and strong.
Why hybrid avoided the career tax
What changed between the two studies was not worker quality. It was design. The authors of the 2024 study, writing for a management audience, point to three implementation conditions: rigorous performance management, coordinated team schedules, and visible support from top leadership (Bloom, Han and Liang, 2024, HBR). The first condition makes work legible without hallway sightings. The second preserves collaboration by having teams in the office on the same days. The third removes the shadow negotiation over whether remote days are career-safe.
Read that list as an operating system, not a policy. If output is defined tightly and reviewed regularly, if teams plan their in-person time for work that benefits from it, and if leaders back the system in public, then the mechanisms that drove the early promotion penalty lose their force. Hybrid did not magically create fairness. It engineered it.
Async is the mechanism that makes work legible
Async is often sold as convenience. It is better understood as the discipline that turns effort into evidence. GitLab’s public handbook describes a “handbook-first” model in which communication is documented so that answers can be self-served. They state plainly that asynchronous collaboration is impossible without strong documentation, since follow-up questions otherwise block work for hours or days. Their handbook runs to more than 2,000 printed pages, a sign of how much written infrastructure async requires (GitLab handbook, 2026). This is not a randomized trial. It is a practitioner case, but it articulates the mechanism that the 2024 hybrid study implicitly depended on: make contributions auditable in writing, and managers can evaluate output rather than presence.
Matt Mullenweg, the chief executive of Automattic, offers a practitioner taxonomy that fits the same logic. His “Five Levels of Autonomy” frames Level 4 as asynchronous work evaluated by output, not by hours or process. Level 5, he argues, would surpass any co-located firm, a strong claim without empirical backing in that essay, but the core definition of async-as-output is useful as design intent, not as proof of superiority (Mullenweg, 2020). The evidence we do have shows that when hybrid is built on clear performance management and coordinated norms, promotion gaps do not appear, even as attrition improves (Bloom, Han and Liang, 2024).
The practical read for leaders
If you are holding hybrid at arm’s length because you fear invisible careers, the best evidence says the risk is controllable. It requires managerial systems that do not confuse attendance with achievement, and team practices that let people move without losing context. Four moves stand out:
- Publish the work. Define outputs, owners, and due dates in shared documents. Insist that decisions and rationales land in writing where they can be found later. This is the spine of an async model and the antidote to presence bias (GitLab handbook, 2026).
- Tie reviews to artifacts. Calibrate performance conversations and promotion cases to the documented body of work, not to anecdotes about who spoke most in meetings. This aligns with the “rigorous performance management” condition that supported hybrid without harming careers (Bloom, Han and Liang, 2024, HBR).
- Coordinate office time at the team level. Use in-person days for work that benefits from co-location, and schedule them so colleagues overlap by design. Trip.com did this as part of its experimental setup, which helped hybrid perform without career costs (Bloom, Han and Liang, 2024; Bloom, Han and Liang, 2024, HBR).
- Test and measure. Run A/B tests on scheduling and review practices, as the Trip.com authors recommend, then track attrition, performance, and promotion outcomes. Their internal estimate is that lower turnover saved millions in recruiting and training costs, a profit lever boards notice. Treat that figure as the company’s own estimate, but the direction of travel is clear (Bloom, Han and Liang, 2024, HBR).
The early remote trial taught a hard truth: if you remove visibility without redesigning how work is judged, you risk hurting careers even when output rises (Bloom, Liang, Roberts and Ying, 2015). The latest hybrid trial shows the inverse: if you redesign the system so that output is visible and schedules are coordinated, you can cut quits without denting performance or promotion (Bloom, Han and Liang, 2024). Async is not a perk. It is the operating discipline that makes those results durable. Build it, and the promotion penalty becomes a choice, not a fate.
Sources
- Bloom, Han and Liang (2024). Hybrid working from home improves retention without damaging performance, Nature. A preregistered randomized trial at Trip.com where two days at home cut attrition by a third with no loss in performance or promotions, grounding the claim that hybrid need not harm careers. 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 experiment that specifies three conditions for successful hybrid and notes the company’s estimated savings 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 at Ctrip where fully remote workers were more productive with lower attrition but had lower promotion rates conditional on performance, illustrating the visibility risk. https://academic.oup.com/qje/article-abstract/130/1/165/2337855
- GitLab (2026). GitLab — handbook-first & asynchronous work (all-remote handbook). Practitioner case asserting that async requires strong documentation so contributions and decisions are accessible without meetings, the mechanism that makes output legible. https://handbook.gitlab.com/handbook/company/culture/all-remote/asynchronous/
- Mullenweg (2020). Five Levels of Autonomy. Practitioner framework defining async at Level 4 as output-based work, used interpretively to describe design intent rather than to claim outcome effects. https://ma.tt/2020/04/five-levels-of-autonomy/