In a colocated office documentation is a convenience. In a distributed team it carries load that the office used to carry invisibly, which changes what it must contain.

An office broadcasts context continuously

People in a shared space absorb a large amount of information they never sought: what other teams are working on, which problems are urgent, who is under pressure.

None of that is written down because none of it needs to be. It arrives through proximity and it is refreshed constantly at no cost to anyone.

Remove the shared space and that channel closes entirely. The information still exists, but now it exists only in individual heads and private conversations.

Written records become the only shared surface

A distributed team's collective understanding is whatever is legible to all of it. Anything held in one person's head is available only when that person is awake and asked.

This makes documents load-bearing in a way they were not before. A missing decision record is not untidiness, it is a gap in what the team knows.

It also means the value of writing something down scales with the number of people who would otherwise have to ask, which grows as the team spreads.

Decisions matter more than descriptions

Most documentation effort goes into describing how things currently work, which is the part people can usually discover by inspection if they have to.

What cannot be recovered by inspection is why a thing works that way: what was considered, what was rejected, and what constraint made the choice necessary.

Teams that record decisions and their reasoning save far more time than teams that maintain elaborate descriptions, because the reasoning is the part that keeps getting relitigated.

Maintenance is the part that fails

Documentation decays because the cost of updating falls on whoever changes the system and the benefit falls on whoever reads it later. Those are rarely the same person.

Without a mechanism that ties the update to the change, the written record drifts from reality and then loses credibility, after which people stop consulting it.

Once trust is gone the team reverts to asking individuals, and the distributed structure starts paying the full cost of its missing ambient channel.

Volume is not the objective

Teams that respond to this by writing everything create a corpus too large to search and too stale to trust, which fails for a different reason.

The useful target is a small set of documents covering the questions that get asked repeatedly, kept current because they are consulted often enough for errors to be noticed.

Everything else can live in the flow of work, where it is cheap to produce and where nobody expects it to remain accurate indefinitely.