A First Checkpoint: Flexibility Is Not the Same as Isolation

Remote work now has a clearer shape than the usual office-versus-home argument suggests. The randomized trial showed that a small, coordinated return to shared space can improve measured performance and retention in a fully remote team. The important judgment at this checkpoint is more precise: flexibility can be preserved, but coordination cannot be left to chance. The next question is whether remote work also changes access to opportunity beyond the firm—and that question requires a different unit of analysis and a different standard of evidence.

The strongest conclusion from the monthly-office-day trial

The trial supports a narrow causal claim. Among eligible customer-service employees who volunteered and could reach the office, assignment to work together one fixed day per month for nine months improved later outcomes relative to assignment to remain fully remote. The treatment group eventually handled more calls per hour, did not show a decline in the measured customer and audit-quality outcomes, and experienced lower cumulative attrition. These effects remained visible after the scheduled visits ended.

That conclusion is stronger than a before-and-after story because the control group continued under the same remote operating model while the treatment group received a coordinated opportunity for contact. The comparison was made within a standardized task environment, with fixed teams and centrally routed calls. It is therefore reasonable to attribute the difference in the measured outcomes to the assigned policy, subject to the sample and setting boundaries.

It is not reasonable to turn that result into a general verdict that offices are necessary. The intervention was not a return to the old workplace. Employees worked remotely between visits. The policy changed the timing and density of contact while leaving pay, hours, technology, formal training, and call routing unchanged. Its causal lesson is about a particular coordination design: limited, shared presence can matter even when it occupies only a small fraction of working time.

The trial also tells us what it does not isolate. Productivity rose gradually, communication patterns changed, managers were reported to give more regular feedback, and attrition fell. Those findings fit a combination of peer learning, managerial input, stronger ties, and attachment to the firm. But the study did not observe every interaction directly or assign employees to separate “feedback,” “peer learning,” and “social connection” treatments. The randomized result is strongest at the level of the policy and its outcomes; the explanation remains a set of mutually compatible mechanisms.

That distinction matters for application. A team cannot copy the headline by adding a calendar event and assuming the same result will follow. The tested treatment created overlap: people were present together, sat near one another, shared breaks, and could reach managers in real time. It also supplied transport, meals, and a workable office. If a proposed office requirement creates little common time, imposes a large travel burden, or does not connect people who need one another, it is not the same intervention. The relevant question is whether the chosen design restores a constrained function.

This also changes how a result should be diagnosed.