The debate over hybrid work is usually framed as a fight about location. It is really a fight about management. When a firm sends people home two days a week without changing anything else about how it runs, it does not get the results that made headlines. It gets a diluted version of the office, with the same instinct to check whether people are online. The evidence is clear that hybrid can work. It is equally clear about the conditions under which it works, and those conditions are easy to skip.
The result had a recipe
The randomized trial at Trip.com is the reason hybrid has a serious evidence base. More than 1,600 graduate employees were randomly assigned to two days at home or five in the office, and resignations among the hybrid group fell by about a third with no measured drop in performance or promotion (Bloom, Han and Liang, 2024). What is often lost in the retelling is that the authors do not present this as an automatic property of working from home. Writing in Harvard Business Review, they name three implementation conditions the outcome rested on: rigorous performance management, coordinated team or company-level schedules, and visible support from senior leadership (Bloom, Han and Liang, 2024, Harvard Business Review).
Read that list again and notice what it is not. It is not surveillance software, not activity monitors, not mandatory camera-on meetings. Every item is about defining and coordinating work rather than watching it. The schedules were set so teams overlapped on the same days. Performance was assessed on results the firm could actually measure, which is why the equivalence tests on review scores meant anything (Bloom, Han and Liang, 2024). Leadership backed the design instead of quietly signaling that the office was where careers were really made.
What happens when the conditions are missing
Remove those supports and a predictable thing happens. A manager who cannot see the work and has no agreed measure of it falls back on the only signal left, which is availability. Green status lights become the metric. Meetings multiply because a meeting is the manager's proof that work is occurring. This is presence-based management with the desk taken away, and it is strictly worse than the office it replaced, because now the watching happens through a webcam.
The all-remote software company GitLab describes the alternative discipline in operational terms. It runs handbook-first, documenting communication in a public handbook that would run past 2,000 printed pages, and it states plainly that asynchronous work is impossible without strong documentation, because otherwise a single follow-up question blocks a colleague for hours or days (GitLab handbook, 2026). That is a self-declared account rather than a measured outcome, and it should be read as one. But the logic tracks the Trip.com conditions exactly. Documentation is coordination made durable. It lets a colleague self-serve an answer instead of pinging someone to prove they are at their desk. The written record substitutes for the manager's line of sight.
Coordination is the thing, not location
The deeper lesson is that the office was never the source of coordination, only one clumsy way of buying it. Proximity gives you overlap and quick answers by accident. A managed hybrid model has to produce them on purpose, through shared schedules and a shared written record. When firms mandate return to office in the name of collaboration, they are often reaching for the accidental version because building the deliberate version is work. The Trip.com authors are explicit that the deliberate version is what delivered the retention gain and protected performance (Bloom, Han and Liang, 2024, Harvard Business Review).
A caution belongs here. The trial covered graduate employees at a single Chinese technology firm, so the specific numbers are not a universal constant. And GitLab's handbook culture is a case study the company tells about itself, with reasons to present it favorably. Neither weakens the structural point. The gain from hybrid is not a gift of geography. It is earned by measurement and coordination, and it evaporates without them.
The practical read
Before a leadership team debates how many days at home, it should answer three prior questions. Can we define good work in this role well enough to measure it without watching someone do it? Do our teams share enough schedule overlap to unblock each other? Will senior leaders back the arrangement rather than reward the people who show up? A firm that cannot answer yes is not choosing between office and hybrid. It is choosing between surveillance in the building and surveillance over video. The design is the whole point, and it is the part most mandates quietly leave out.
There is a reason the deliberate version gets skipped. Writing down how work is coordinated is slow, unglamorous, and easy to defer, while ordering people back to their desks is fast and feels decisive. But the trial did not reward the decisive gesture. It rewarded the coordination, and it produced its retention gain without any of the monitoring that presence-based managers reach for out of habit (Bloom, Han and Liang, 2024). A leader who wants the result the study describes has to want the recipe that produced it, not just the outcome it reported.
Sources
- Bloom, Han and Liang (2024). Hybrid working from home improves retention without damaging performance (Nature 630). The RCT establishes that managed hybrid cut quits by a third without harming performance, the outcome whose preconditions the article dissects. 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). Names the three implementation conditions (rigorous performance management, coordinated schedules, leadership support) that separate managed hybrid from surveillance. https://hbr.org/2024/10/one-company-a-b-tested-hybrid-work-heres-what-they-found
- GitLab handbook (2026). Handbook-first and asynchronous work. Self-reported case showing documentation as the coordination substrate that replaces line-of-sight, supporting the anti-surveillance argument. https://handbook.gitlab.com/handbook/company/culture/all-remote/asynchronous/