Both sides of this argument are overconfident.
The pro-office case usually rests on culture, collaboration and a vague sense that things were better before — with the causal claim assumed rather than demonstrated. The pro-remote case leans on studies with narrow populations and self-reported output, then generalises hard.
Having spent a lot of time looking at what work patterns actually show, I think the honest position is less satisfying than either: office presence has no consistent measurable effect on knowledge-work productivity, and it does have effects on things that are not productivity but still matter.
That is worth unpacking, because it changes what a sensible policy looks like.
What the evidence does not support
There is no reliable finding that knowledge workers produce more output in an office. Results are mixed across studies, effect sizes are small where they exist, and the measurement is genuinely hard — most of what gets counted as productivity in this literature is either self-reported or a proxy that breaks under scrutiny.
Anyone telling you the data clearly supports RTO for productivity reasons has not read much of it. That includes people whose conclusion I happen to share on other grounds.
Equally, anyone telling you remote work is unambiguously more productive is doing the same thing in the other direction. The pandemic-era studies that get cited most were run under conditions nobody would design deliberately, and the populations were rarely representative.
What the evidence does support
Three things show up more consistently, and none of them is productivity.
Onboarding is harder remotely. New joiners ramp more slowly without ambient exposure to how the team actually works. This one has the strongest support and the most obvious mechanism.
Weak ties decay. The relationships you have with people you do not work with directly weaken without incidental contact. Those ties are how information crosses team boundaries, so the effect is real even though it is invisible quarter to quarter.
Coordination costs rise across time zones. Not remote per se — distributed. A remote team in one time zone has very different coordination properties from one spread across three, which the time zone playbook goes into.
Notice that all three respond to design rather than requiring five days in a building. Structured onboarding with an assigned buddy. Deliberate cross-team forums. Periodic in-person gatherings with actual agendas. These are cheaper than a mandate and they target the mechanism instead of the proxy.
The thing mandates are usually solving
Microsoft's Work Trend Index put a number on the real driver: 87% of employees said they were productive, while 85% of leaders said hybrid work made it hard to be confident that they were.
That gap is not a productivity problem. It is a measurement problem experienced as anxiety.
A return-to-office mandate resolves it, in the sense that seeing people at desks restores the leader's confidence. It does not resolve it in the sense of producing evidence. You have swapped an uncomfortable uncertainty for a comfortable proxy, and the proxy — physical presence — was never a good measure of knowledge work even when everyone was in the building.
We build measurement tooling, so treat this as an interested opinion: I think a lot of RTO policy is leaders buying back a feeling of visibility, and I think it is an expensive way to buy it.
If you are going to mandate, do it properly
Not an argument against office policy. An argument against unfalsifiable office policy.
Decide what you think will improve. Write it down before the policy starts. Measure it for a quarter beforehand — onboarding ramp time, cross-team collaboration, focus time, whatever you claimed the problem was. Then measure the same thing for a quarter after.
Almost nobody does this, which is why the debate never resolves and why every company relitigates it annually. Both sides are arguing from anecdote because neither took a baseline.
And commit in advance to what you will do if the numbers do not move. That is the part that makes it a real experiment rather than a decision with a measurement ritual attached.
What we would actually suggest
Be specific about the problem. "Collaboration" is not a problem statement; "new engineers are taking five months to become productive and it used to be three" is.
Then match the intervention to it. Slow onboarding gets a redesigned onboarding programme, possibly with in-person time in the first month. Weak cross-team ties get structured forums and a quarterly gathering. Low focus time — which is the most common finding when teams actually measure — gets fixed by attacking meeting load and interrupts, and an office typically makes it worse rather than better.
The general method is in measuring remote productivity without surveillance, and the tooling question in the remote monitoring guide.
The uncomfortable summary
If your team is distributed and struggling, the office might help with some of it. If your team is distributed and you simply cannot see what is happening, the office will make you feel better without telling you anything you did not already have available.
The second case is far more common, and it is cheaper to fix with a baseline than with a building.
ProdView measures focus time, meeting load and working patterns from activity metadata, which is what a before-and-after on any office policy actually requires. Free for three seats — enough to take the baseline before you announce anything.