There is an irony sitting at the centre of distributed work that nobody planned for. The companies that gave up their offices, the ones that made “we are fully remote” a recruiting line, are now spending more per employee on getting people into the same room than many of the companies that never left the office at all.

It sounds like a contradiction. It is closer to an accounting inevitability. Remove the office and you remove the default mechanism by which colleagues encounter each other, which means every encounter now has to be deliberately purchased.

The offsite replaced the office, and it is not cheap

A traditional office provides an enormous amount of unplanned interaction as a byproduct of rent. Nobody budgets for the corridor conversation, the overheard call, the lunch that resolves a two-week misunderstanding. Those things arrive free with the lease.

Distributed companies discovered, usually around eighteen months in, that these interactions were not decorative. They were doing structural work: transmitting context to new hires, surfacing problems early, building the trust that makes disagreement productive rather than corrosive.

Replacing them requires deliberate, scheduled, expensive gatherings. A typical fully distributed company of two hundred people running two annual all-company gatherings will spend a meaningful multiple of what an equivalent office-based company spends on team events, because the gathering is not a perk sitting on top of the office. It is the office, compressed into four days.

The comparison that matters is not gathering cost versus zero. It is gathering cost versus rent, and by that measure remote companies are usually still ahead.

What distributed teams are actually buying

The mistake most companies make on their first attempt is treating the gathering as a conference. Presentations, a keynote, breakout sessions, a schedule packed to the last minute.

This gets it backwards. Information transfer is the one thing distributed teams are already good at. They have documentation, recorded video, asynchronous written updates. A slide deck delivered in a hotel ballroom is a worse version of something they could read at home.

What they cannot do remotely is the unstructured part: the informal conversation, the shared meal, the accidental encounter between two people from different functions who turn out to have overlapping problems. Those are the scarce goods, and a schedule packed with sessions systematically crowds them out.

The gatherings that work tend to be deliberately underprogrammed, with long meals, generous gaps and a small number of structured moments rather than many. That is harder to organise than it sounds, because empty space in an agenda looks like waste to whoever is approving the budget.

It is also why a growing number of distributed companies hand the logistics to specialist event planning companies rather than assigning them to an operations manager as a side project. The work is not intellectually difficult, but it is enormously time-consuming and highly failure-sensitive, and the internal cost of doing it badly tends to exceed the external cost of doing it properly.

The cost comparison nobody runs

Ask a remote company what its gatherings cost and you will usually get the direct programme figure: flights, accommodation, venue, catering. That number is incomplete in both directions.

On the cost side it omits the internal time. Someone spends weeks organising it. Everyone spends days travelling to it. For a four-day gathering, the loaded salary cost of attendance frequently exceeds the entire logistics budget.

On the benefit side it omits the counterfactual. What does it cost to have a distributed team that has never met? The available evidence points at slower onboarding, weaker cross-functional collaboration and higher voluntary attrition, particularly among employees hired remotely who never developed relationships beyond their immediate team.

Neither figure is precise. But a company that has run neither calculation is not making a decision about gatherings, it is expressing a preference about them.

Frequency matters more than scale

If there is one finding that recurs across companies that have iterated on this for several years, it is that two modest gatherings beat one impressive one.

The reason is decay. Whatever a gathering produces in terms of relationships and shared context erodes over the following months. A single annual event means eight or nine months of decay before the next reset. Two events halve that window at considerably less than double the cost, because smaller gatherings scale more efficiently than the headline numbers suggest.

The corollary is that grandeur is usually a poor investment. The memorable part of a company gathering is almost never the production values. It is who someone happened to sit next to at dinner.

What tends to go wrong

Scheduling every hour. Covered above, and it remains the single most common failure.

Inviting everyone to everything. Two hundred people in one room is a broadcast. Twenty people in one room is a conversation. Good gatherings alternate between the two rather than defaulting to the first.

Ignoring the arrival and departure days. People arrive tired and leave distracted. Programming substantive content into those windows wastes it.

Treating it as a reward. Framing the gathering as a perk creates the expectation of a holiday and the resentment that follows when it turns out to involve work. Framing it as work with unusually good food sets expectations correctly.

Skipping the follow-up. Whatever gets decided in the room evaporates within a fortnight unless someone writes it down and assigns it.

The office was never the point

The wider lesson here is not really about events. It is that the office was always bundling two separate products: a place to do focused work, and a mechanism for maintaining a social organism.

Distributed work unbundled them, and it turned out the first product was easy to replace and the second was not. Companies that acknowledged this early and started paying explicitly for the second one have generally done well. Companies that assumed it would sort itself out have generally not.

The spending is not evidence that remote work failed. It is evidence that it is being taken seriously enough to be resourced properly, which is more than could be said for most experiments in how people work together.