Add a person to a team of four and you have not added a fifth of the capacity. You have added six new pairs of people who now have to agree with each other.
Communication cost grows with the square of the team. Output does not.
This is old and well documented, and it keeps being rediscovered because the fix is unpopular: to go faster, you usually have to take people off the project rather than add them.
Where the time actually goes
Small team
- Decision made at the desk
- Context already shared
- One person owns the outcome
- Correction costs an afternoon
Large team
- Decision needs a meeting
- Context has to be rebuilt
- Ownership is distributed
- Correction costs a sprint
Great things in business are never done by one person. They are done by a team of people — a small one.
Adapted, and the second half matters
What actually makes them fast
- Shared context. Nobody has to be briefed, because everybody was there.
- Reversible decisions. Small teams pick the door they can walk back through, and walk through it sooner.
- Nobody to hand off to. The person who found the problem fixes the problem.
- Short feedback. A mistake surfaces in hours, not at the review.
The limits are real and worth naming. Small teams carry a bus factor of roughly one, they are terrible at anything that needs sustained coverage, and the shared context that makes them fast is exactly what makes onboarding a new person slow.
The trade you are actually making
Speed now, in exchange for resilience later. That is a fine trade for a first version and a bad one for a system that has to run for ten years. The mistake is not choosing one — it is choosing one and pretending you did not.
If you have to grow
Split before you scale. Two teams of four, each owning something end to end, keep most of the speed. One team of eight keeps almost none of it.
Does this apply outside software?
The communication maths does. The rest depends on how divisible the work is: a kitchen and a newsroom both hit the same wall, and both solve it the same way — smaller units, each owning a whole thing.
The bus factor point is the honest counterweight. Most articles like this leave it out.
“Split before you scale” is going on our wall.