In 1967 a programmer named Melvin Conway sent a paper to the Harvard Business Review. It was called “How Do Committees Invent?” The editors rejected it. By Conway’s account, they said he had not proved his thesis.
They had a point. His evidence was a handful of examples. A firm with eight programmers had put five on a COBOL compiler and three on an ALGOL compiler. The COBOL compiler ran in five phases; the ALGOL compiler ran in three. Two military services told to build a common weapon system had, “after great effort,” produced “a copy of their organization chart.”
Conway sent the paper to Datamation, the big computer magazine of the day, which printed it in April 1968. The main claim sat near the end. Forty-two years later Conway noted, with some amusement, that it had reached print “with its thesis statement in the third-last paragraph,” after more than 40 paragraphs of build-up. The claim: organizations that design systems produce designs that copy their own communication structures.
The idea might have stayed a curiosity. Then Fred Brooks quoted it in The Mythical Man-Month and called it “Conway’s Law.” The name stuck.
The empirical test came much later. In the 2000s, three researchers at Harvard Business School, Alan MacCormack, John Rusnak and Carliss Baldwin, set out to test what they called the mirroring hypothesis. They compared pairs of software products that did similar jobs, one built by a tightly knit company team and one by a loosely connected group such as an open-source project.
“In all of the pairs we examine,” they reported, “the product developed by the loosely-coupled organization is significantly more modular.” The difference in how far a change could ripple through the code reached up to a factor of eight.
The working paper went out under the name of Harvard Business School.