Set the Conditions, Then Pick the Location

Six pieces of evidence, briefly, before the audit. Two randomized trials asked whether working from home works and reached opposite answers. Every negative finding in that literature reduces to a condition degrading rather than a location failing. Coordination time rose while uninterrupted hours shrank. Collaboration networks went static and siloed. Junior people stopped receiving feedback. And in the offices people are nostalgic for, none of those conditions had been built. They were being propped up by proximity accidents nobody designed.

Which means the useful question was never where people should sit. It is which condition is currently binding, because that answer decides everything downstream including the location, and because it is an answer your own organization can produce in about an afternoon.

Question one, attention

Take one person who does making work. Count last week's uninterrupted blocks of ninety minutes or more. Not blocks with nothing scheduled in them. Blocks with no meeting, no expected response and no queue waiting.

Zero or one is the common answer and it is a finding, not a scheduling accident. The interruption research is clear that this cost does not show up where people look for it. In a controlled experiment, interrupted tasks were finished in less time with no drop in quality, because people compensated by working faster, and the price landed on stress, frustration and effort. Gibbs and colleagues found the same shape at firm scale, with total hours up about 30 percent and average output unchanged.

What it means for location. Nothing, directly. Perlow found this exact shortage in a co-located software team in 1999 and repaired it by changing how the group used time rather than by moving anyone. If your answer is zero, a mandate will not change it, because a mandate does not delete a single recurring invitation.

Question two, alignment

Take two people on different teams who depend on each other. Ask each of them separately what the current priority is and what the other team is working on. Compare the two answers without softening either.

The distance between the answers is the alignment gap, and it is the only measurement of it that a meeting cannot flatter. Alignment is not agreement. Mathieu and colleagues showed that convergence in shared mental models predicts better team processes and that better processes predict better performance, so the gap is upstream of everything you would otherwise try to fix.

What it means for location. This is the condition proximity helped most and helped most narrowly. The engineering communication research puts the effect of adjacency at around fifty meters, so an office restores the ties between people already seated together and does little for the gap between two functions on different floors. If the pair you just tested sit far apart, the building was never closing that gap either.

Question three, authority

Take one decision that went up the chain last month. Name what the approver knew that the requester did not. If the honest answer is nothing, that decision was sitting at the wrong level.

Then run the harder version. Ask five leaders independently who can approve an unplanned five thousand dollar spend, who can move a delivery date, and who can stop a piece of work that is no longer worth doing. Any spread in the answers is a decision your organization has not made, currently being resolved by whoever is most available.

What it means for location. Formal authority is the right to decide. Real authority is who actually decides, and it follows the information. An office lets real authority route itself informally, which is why an organization can run for years on decision rights nobody ever wrote down. If bringing people back made decisions faster, that is not evidence the office works. It is a measurement of how much of your operating model lives in a corridor.

Question four, adaptability

Name one thing the team learned this quarter that changed a rule, a routine or a decision. Not a lesson that was captured. One that changed something. If nothing comes to mind, that is the finding.

Then ask the version that has a number attached. How much feedback are your least experienced people actually receiving on their work? Emanuel, Harrington and Pallais measured exactly that and found proximity raised it by 18.3 percent for junior engineers, and a separate study found remote work degraded quality particularly for inexperienced workers and lowered promotion rates.

What it means for location. This is the strongest case the office has, and it is narrow and specific. It is about novices acquiring judgment rather than about experts executing codified work, which is why patent examiners at the USPTO produced more output when allowed to work from anywhere. If this is your weakest condition, proximity genuinely helps, and it should be aimed at the people who are learning rather than distributed evenly across everybody.

The decision rule

Whichever of the four came back worst is the constraint costing you performance today. Fix that condition. Then, and only then, choose a location policy, and choose it as a dose aimed at the specific people the research says benefit.

The dose is not a matter of taste. The field experiment that randomized how many days people worked from the office found around two days produced the best reported work-life balance, job satisfaction and connection with no difference in performance ratings. The randomized trial of 1,612 people at Trip.com found two days a week at home left performance grades unchanged and cut quit rates by a third. Two experiments, different countries, both interior.

If the conditions are built, the team performs distributed, hybrid or crammed onto an open plan floor. If they are not, the office helps, and what you have learned is what the office was carrying.

The two ways to get this wrong

Mandating presence to repair a condition. This works, in the sense that the symptom improves. It works by reinstating a workaround, it leaves the condition exactly as broken as it was, and it charges everybody a commute in perpetuity to keep the workaround running. It also has a ceiling at about fifty meters.

Committing to fully distributed work without building anything. Also popular, also cheap, also an intervention with no stated endpoint. An organization that never deliberately built the four conditions and then removes the thing that was accidentally holding them up has not chosen flexibility. It has chosen to find out.

Both are the same error in different clothes. Somebody picked a location policy, applied it to everybody, and read the resulting compliance as a result.

What this audit is and is not

It is four readings taken from one vantage point, on one week, by people inside the situation they are describing. That makes it a starting hypothesis rather than a verdict, and it is worth arguing with where your own judgment disagrees. The value is that it moves the conversation off a preference and onto something checkable, which is more than most organizations have when they announce a policy.

If the answers are close together, or if the argument in your organization has already hardened past the point where four questions will move it, a structured reading with more than one respondent is the next step rather than a longer memo.