The remote work conversation has been remarkably unproductive, largely because it's framed as a binary with two camps asserting incompatible things about productivity.
The research is more nuanced and points somewhere neither camp finds convenient.
What the studies actually find
Results vary substantially by task type, which is the finding that gets lost.
For individual work with clear outputs — writing, coding, analysis, processing — studies have generally found remote work neutral to positive on productivity, with fewer interruptions cited as a major factor.
For collaborative work requiring rapid iteration, and particularly for work involving ambiguity where people need to develop shared understanding, results are less favourable.
For learning and development, particularly for junior people, the evidence is fairly consistently negative. A lot of skill acquisition happens through observation and incidental interaction, and those don't survive the transition well.
So the honest summary is that it depends on the work, which is unsatisfying as a policy input and is what the evidence supports.
The communication network finding
One of the more interesting results concerns how collaboration networks change.
Research on large organisations moving to remote work has found that networks became more siloed — people communicated more intensively within their existing close groups and less across the organisation.
Weak ties, the connections between different parts of an organisation, declined. That's significant because weak ties are disproportionately associated with novel information and with innovation, going back to some well-established sociology.
The mechanism is straightforward: casual cross-group contact is largely incidental, and incidental contact requires physical proximity or deliberate replacement. Scheduled meetings replace planned interaction, not unplanned interaction.
Hybrid is not a compromise
The most common arrangement is some days in and some at home, usually presented as splitting the difference.
Done badly it delivers the disadvantages of both. If people attend on different days, the office is populated but not with the people you need, and you take video calls from a desk in an open-plan room — which is worse than either alternative.
The evidence favours coordinated attendance: everyone in on the same days, with those days structured around the activities that genuinely benefit from being together. Collaborative sessions, planning, onboarding, the social contact that maintains relationships.
And crucially, not scheduling individual focused work on office days. If people come in to sit silently with headphones on, the commute bought nothing.
The measurement problem
Both sides of this argument cite productivity data, and most of it is weak.
Output is hard to measure for knowledge work. Proxies get used — activity metrics, hours logged, messages sent — and these measure visibility rather than value. Software that monitors keystrokes and screen time is measuring compliance, and there's reasonable evidence it degrades trust and therefore performance.
Self-reported productivity, used in many studies, is unreliable in both directions. People report being more productive at home; managers report the opposite about the same people. Both are reporting perceptions shaped by what they can observe.
The measures with the best claim are outcome-based — did the thing get delivered, at what quality, by when — and those are the hardest to standardise across roles.
What's genuinely different
A few effects seem robust regardless of the productivity question.
Commuting time. Straightforwardly returned to the employee. For anyone with a long commute this is a substantial change in quality of life and it's the single largest factor in stated preference.
Geographic hiring range. Access to candidates outside commuting distance, which matters enormously for specialised roles and for organisations outside major cities.
Inclusion effects. Remote work has clear benefits for people with caring responsibilities, disabilities, or health conditions that make commuting difficult. This is one of the strongest arguments and it's frequently underweighted.
Proximity bias. The documented tendency for people who are physically present to receive more recognition and advancement. In hybrid arrangements this creates a real equity problem, and it's the strongest argument for coordinated rather than individual choice.
What I'd actually ask
Rather than debating location, the more useful questions are about the work.
Which activities genuinely require synchronous interaction, and which have simply always been done that way? Most meetings fail this test.
How do junior people learn here, and does that mechanism survive distribution? If the answer is by watching, it doesn't, and something has to replace it deliberately.
How does information reach people who aren't in the room where it's discussed? If the answer is that it doesn't, you have a documentation problem that remote work exposes rather than causes.
Those questions produce actionable answers. "Office or home" produces an argument.
The onboarding gap
If there is one area where the evidence points most clearly, it is bringing new people in. Distributed onboarding is measurably harder and most organisations underinvest in it.
The mechanism is that a new joiner in an office absorbs an enormous amount incidentally — how decisions get made, who to ask, what the unwritten norms are, what tone is normal. None of that is documented anywhere and none of it survives distribution.
What replaces it has to be deliberate: assigned buddies, structured introductions to people outside the immediate team, written explanations of things nobody thought needed writing, and considerably more contact in the first months than feels necessary.
Organisations that do this well report no particular disadvantage. Organisations that assume onboarding happens by itself lose people in the first year at noticeably higher rates.