The Commons You Can't Buy.

In 2013, Google made a bet that looked unloseable on paper. The Moto X smartphone would be assembled not in Shenzhen but in Fort Worth, Texas — cheaper logistics, "Made in the USA" on the box, Flextronics running the line. Within a year the factory was closed. The machines worked. The instructions were followed. What Texas could not supply was what Shenzhen has in absurd abundance: a dense web of component suppliers, toolmakers and engineers within an hour's drive, where someone always "knows someone whose cousin can supply several hundred thousand of a needed component by morning" (Wang, 2025; Shih, 2014).

The plant failed for a reason that appeared on no spreadsheet: the knowledge it needed did not live inside its four walls.

The industrial commons: knowledge between firms

Harvard's Gary Pisano and Willy Shih named this back in 2009: the industrial commons — the shared pool of process-development skills, engineering expertise, supplier networks and skilled trades that no single firm owns but every firm draws on. Their unsettling finding was a vicious cycle: when work is offshored, it is not just factories that leave. The apprenticeships stop. The local toolmaker closes. The next generation never builds the know-how — and it becomes "cheaper" to send the next process away too.

Dan Wang's Breakneck shows the same loop running in reverse. China's manufacturing depth is a story of clustering: more than 500 towns specialise in specific products, and process knowledge travels in engineers' heads from one firm to the next, enriching the whole. Henry Farrell found the identical mechanism in Bologna's "Packaging Valley" — know-how as arcane as how to reliably staple the thread onto a tea bag.

Same mechanism, opposite signs. The commons compounds, or it erodes. It never stands still.

Why this is your problem, specifically

If you run a brownfield plant in South Africa — or Manchester, or Milwaukee — you are living inside the erosion spiral: the artisan who could hear a bearing failing retires with no apprentice behind him; the local machine shop that once turned a shaft overnight closed in 2019; every serious fault escalates to an OEM engineer on another continent, so your own people never build the diagnostic muscle.

A machine-learning model can flag an anomaly at 2am; it cannot machine the part, and it cannot conjure the artisan who knows which of three plausible causes is the real one. AI raises the return on process knowledge — and everywhere the commons has thinned, there is less of it to return on. Africa's manufacturing "missing middle" sharpens this: AfCFTA opens the borders; it cannot, by itself, build the commons. That happens plant by plant, town by town — which means it can start with you.

The systems view: your plant needs a radar

In Stafford Beer's Viable System Model, System 4 is the function that looks outward and forward. Most brownfield plants have magnificent System 3s (running today) and no System 4 whatsoever — nobody maps which local suppliers survive, visits the TVET college, or knows what the plant across the road is doing about the same skills cliff.

Your plant's most dangerous blind spot is not on the shop floor. It is the assumption that the factory gate is the boundary of the system.

The Digital Kaizen Homeostat makes the point cybernetically: the self-correcting loop between people, process and technology does not stop at the fence. A homeostat that only senses inside the plant keeps "correcting" toward a local optimum while the ecosystem it depends on disappears.

What works: feeding the cluster

• Toyota's supplier learning networks. Toyota does not merely audit suppliers; it teaches them — seconding engineers into supplier plants and running joint kaizen — deliberately building process knowledge it does not own.
• Bologna's spin-out culture. Packaging Valley grew because leaving your employer to start a competing machine shop was treated as the system working. People circulating is the commons breathing.
• One plant playing System 4. The most practical move I have seen in South Africa is modest: give your hardest recurring machining problems to a local shop even when an import is marginally cheaper; take two TVET apprentices a year onto real breakdowns; host a quarterly open morning for neighbouring maintenance managers. Three decisions, none heroic — and in five years that plant sits inside a measurably smarter ecosystem no competitor can order from a catalogue.