Melvin Conway formulated it in 1967 and published it the following year, in a paper called How Do Committees Invent?: any organization that designs a system will produce a design whose structure is a copy of the organization's communication structure.

The reason is mundane. Two people who talk easily produce a smooth interface between their parts. Two who need a meeting to coordinate produce a defined, awkward one. Two who do not speak produce a boundary nobody owns. The system's seams end up where the organization's seams are.

The observable version

Look at any product and you can usually read the org chart off it. Four teams, four modules, and the hard integration problems concentrated exactly where the teams report to different managers. A feature that requires two departments to agree will be worse than one that does not, in proportion to how hard the agreement is.

It applies well beyond software. A company with separate sales and delivery functions produces a customer experience with a seam at handover. A firm where finance and operations do not speak produces reports that do not reconcile.

Using it deliberately

The productive version is the inverse maneuver: arrange the teams the way you want the system arranged.

If you want a product with three cleanly separated parts, build three teams with clear interfaces. If you want a seamless customer experience, one team has to own the whole path — which is a large part of why small firms deliver more coherent service than larger ones, and why that coherence is usually the first casualty of growth.

For a service business this is the most useful reading. The client experiences your internal boundaries whether or not you mean them to. Every handoff between people is visible as a seam to the customer, and a handoff between people who do not coordinate well is visible as a bad one.

It pairs directly with Brooks's law: communication structure limits what a group can build, and adding people changes that structure whether or not anyone intends it.