Four round steel modules on square bases stand close together, but only two are linked by a blue cable.

Returning to the office is not returning to collaboration

Attendance can be recorded; collaboration has to happen. Which encounters actually add value, and what evidence would justify reconsidering an office-attendance policy?

Alejandro Mayorga6 min read

A return-to-office policy can be followed without improving what it was meant to address. People arrive on the required days, take their seats, and attend the scheduled meetings. Attendance is recorded; coordination problems may remain where they were. Before celebrating the return, it is worth holding on to the question that prompted the measure: what interaction did we need to restore, between whom, and to make what work possible?

The answer is not contained in a general preference for the office or for remote work. Sharing a physical space can make someone else’s expertise more accessible and create conversations that would be difficult to anticipate on a calendar. It can also interrupt work that requires sustained concentration. These possibilities become more interesting when we stop treating them as absolute properties of a location and start examining for whom, when, and under what conditions their value emerges.

Two findings worth holding together

In a randomized experiment published in Nature in 2024 (opens in a new tab), Nicholas Bloom, Ruobing Han, and James Liang assigned 1,612 employees at Trip.com in China to a hybrid arrangement offering two days a week at home or to five days in the office. During the six-month trial in 2021–2022, resignations were approximately one-third lower in the hybrid group; there was also no observed harm to performance reviews over the following two years. The finding challenges the assumption that less time in the office must impair work. It does not tell us what would happen under an entirely remote arrangement or establish the right schedule for every company: the arrangement studied still involved three office days.

For their part, Natalia Emanuel, Emma Harrington, and Amanda Pallais studied software engineers at a large company from 2019 to 2024. Their research published in The Quarterly Journal of Economics in 2026 (opens in a new tab) uses office closures and subsequent return mandates to compare changes among teams that had previously worked at different degrees of proximity. In this quasi-experimental, nonrandomized design, sitting close to colleagues increased feedback on code and improved its quality, with benefits concentrated among younger or less-tenured engineers. Experienced engineers, by contrast, wrote less code when seated near their colleagues.

That latter finding makes it inadequate to ask only where each individual produces more. If some of the value of proximity comes from the help an experienced person provides to someone else, we must recognize both contributions. Otherwise, we could encourage an exchange that improves the team’s work while penalizing the person who spends time sustaining it. This is not a problem that can be solved by adjusting attendance days: we would need to reconsider what work we recognize and how we evaluate it.

The studies do not present results that we must reconcile by choosing one and discarding the other. They observe different working arrangements and allow us to keep two considerations in view: a hybrid schedule can retain employees without harming measured performance, and proximity can support learning and quality at the cost of some immediate output by those doing the helping. Neither study covers every kind of task. Together, they make it harder to defend a single measure that declares one work arrangement the winner.

What needs to happen when we meet

What we should examine is a working relationship, not merely each person’s location. When design and operations face a shared problem, for instance, what matters is that those who understand the problem can meet those who have the authority to change the process. Bringing both teams into the office could facilitate that encounter; it does not ensure that their schedules overlap or that they have time to reconstruct what happened. If they spend the day in calls with people in other places, the commute has altered one condition without addressing the need that justified it. Making their presence coincide requires deciding which conversation deserves space and which commitments must move to make it possible.

Nearby office workstations divided by blue panels; people are working separately.

Not every valuable interaction can be planned this way. A small question may be asked because someone happens to be nearby and seems available. Turning every exchange into a mandatory meeting would be a strange way to protect that opportunity. Leaders also need to recognize the value of being available, without treating every interruption as collaboration or every spontaneous conversation as wasted time.

A remote-work policy carries equivalent obligations, even if it meets them differently. Deliverables arriving on time does not guarantee access to context or the ability to ask for help. The organization needs to examine where decisions are recorded, how people reach those who can clarify them, and what happens when a newcomer does not yet know whom to ask. Flexibility should not require every new hire to invent a way to reach everyone else. Sometimes there will be good reasons to meet in person; in other cases, better documentation or a change in availability may solve a problem that never depended on distance.

Evaluating any of these choices requires a concrete expectation. Restoring exchange between two departments is different from helping a newcomer become autonomous in their work. In the first case, we would want to know whether shared problems are detected and resolved more effectively; in the second, what support actually reaches the person learning. Attendance records may show that a policy was applied. To understand whether it was useful, we need to observe what we expected it to make possible.

We must also consider how costs are distributed. The same attendance requirement can impose very different commutes and demands on people’s personal arrangements; a remote policy may leave those who need the most support without enough of it. These consequences do not determine the answer automatically, but they belong in its justification. If a service requires on-site coverage, or material constraints limit the alternatives, those conditions need to be part of the agreement—not presented as proof that one arrangement is universally superior. Within those limits, we should still be able to reconsider a policy that fails to deliver what it promised, without treating each adjustment as a defeat in the debate between office and remote work.

Requiring presence when there is a sufficient reason, and protecting distance when it adds value, demands more of leadership than setting a number of days and checking compliance. It asks us to regard one reassuring image with some suspicion: every desk occupied. The decisive question remains what the people who arrived there can do together. A full office can also be a highly visible way of continuing to work apart.