Everything the manager knows about managing, she learned within thirty feet of her team. Reading a room before a difficult meeting. Noticing that a strong performer had gone quiet at lunch. Calibrating a project's health by the hum of a floor. Then the team went hybrid, half her instrument panel went dark, and nobody retrained her; the organization simply assumed that a manager of people would automatically become a manager of distributed people, the way it once assumed a good engineer would automatically become a good manager.
Her private suspicion, that this cannot really work, deserves to be taken seriously, then confronted with what happened when it was tested.
What the experiments actually found
The cleanest evidence comes from randomized trials run by Nicholas Bloom and colleagues. In the Trip.com experiment published in Nature, 1,612 professionals were randomized into hybrid work, two home days a week, or full office presence. Hybrid cut voluntary quits by a third, from 7.2 to 4.8 percent, with no detectable difference in performance reviews or promotions across two years of follow-up. The detail most relevant here is what happened to the managers in the study: Stanford's account of the trial notes that managers who predicted remote work would hurt productivity had reversed their view by the end. Presence-trained intuition made a testable forecast, and the forecast was wrong.
But the earlier Ctrip experiment, from 2015, keeps the story honest. Fully remote call-center workers were 13 percent more productive and quit less, yet their promotion rate, conditional on performance, fell. Out of sight translated into out of the running. That penalty is not an argument against flexibility; it is a description of unmanaged visibility bias. And visibility bias is not the employee's failure. It is the manager's, because the manager owns the evaluation.
This is the actual job change. In presence, evaluation could free-ride on observation. In async, evaluation has to be designed.
Output, written down
The practitioner playbooks converge on two disciplines. GitLab, which runs all-remote with a public handbook that would exceed 2,000 printed pages, states its operating premise bluntly: asynchronous work is impossible without strong documentation, because every undocumented answer blocks someone for hours. Its position is async-first, not async-absolute; incidents, deadlines and relationship-building revert to synchronous. Matt Mullenweg of Automattic frames the destination as an autonomy ladder in which the decisive level is the one where work is evaluated on output, not hours or process. Neither source is a controlled study, and both have an interest in the model's reputation. But their common mechanism, replace observed effort with defined and documented output, is exactly the corrective the Ctrip promotion penalty demands.
There is an older precedent for this substitution. As Ricardo Semler has told it, he collapsed on a factory visit at 25 and was told by the doctor that he was the healthiest sick person he had ever seen, nothing wrong beyond the way he lived. That afternoon he began dismantling the machinery of presence at Semco, the time clock among the first things to go, and the company stopped counting hours and started measuring whether the pump got delivered. Absenteeism fell. The management technology the async era requires was invented before laptops; presence was always a proxy, and proxies can be retired.
The retraining, concretely
For the presence-trained manager, the transition reduces to four practices. Define output per role clearly enough that a person could self-assess on a Friday. Write decisions and context down, because the handbook, not the hallway, is now where alignment lives. Audit promotions and stretch assignments for a proximity pattern; if office-frequent people are advancing faster than their output justifies, the Ctrip result is running unmanaged inside her team. And keep synchronous time for what it is uniquely good at, which is trust, conflict and ambiguity, not status reporting that a document could carry.
None of this retires her old skills. Reading people, sensing trouble early, building safety in a room remain scarce and valuable; the four practices simply stop those skills from being load-bearing infrastructure. The managers who struggle with async are not the ones who loved people too much. They are the ones who confused seeing work with knowing it was happening. The pump either shipped or it did not, and no amount of floor-walking ever changed that.
Sources
- Bloom, N., Han, R. & Liang, J. (2024). Hybrid working from home improves retention without damaging performance. Nature 630. RCT with 1,612 employees: hybrid cut quits by one third (7.2% to 4.8%) with no performance or promotion penalty over two years. https://www.nature.com/articles/s41586-024-07500-2
- Bloom, N., Liang, J., Roberts, J. & Ying, Z. J. (2015). Does Working from Home Work? Evidence from a Chinese Experiment. Quarterly Journal of Economics 130(1). Full-time WFH raised performance 13% but reduced promotion rates conditional on performance. 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. Managers who predicted productivity losses from hybrid reversed their views by the end of the Trip.com experiment. https://news.stanford.edu/stories/2024/06/hybrid-work-is-a-win-win-win-for-companies-workers
- GitLab (2026). All-remote handbook: asynchronous work. Company-stated operating premise that async work is impossible without strong documentation; handbook-first, async-first but not async-absolute. https://handbook.gitlab.com/handbook/company/culture/all-remote/asynchronous/
- Mullenweg, M. (2020). Five Levels of Autonomy. Practitioner framework in which mature distributed work is evaluated on output rather than hours or process. https://ma.tt/2020/04/five-levels-of-autonomy/