Abrigo·Marrón Systems

Apuntes de Campo Field Note

Governance Is Not Government

Say systems thinking in a plant manager's office and watch what happens to his face. He has heard this before. Somebody from corporate came down with a holistic framework, an integrated dashboard, and a new approval layer, and what it produced was four more signatures on a work permit and a six-week wait for a decision his lead operator could have made correctly in ten minutes. He is not being anti-intellectual. He is pattern-matching on experience, and the pattern is real.

The suspicion runs deeper than bad consulting, though. The vocabulary itself sounds collectivist. Whole systems. Collective behavior. Central coordination. The good of the system over the good of the part. It reads like an argument for a larger central authority and a smaller sphere of individual judgment, and anyone who has watched a bureaucracy grow in the name of integration is entitled to be wary.

It does not help that the most famous applied experiment in management cybernetics was a national economic project in Chile in the early 1970s, undertaken for a socialist government and ended by a coup. Stafford Beer built a real-time industrial information network, and the association has clung to the field ever since. Fair enough. It happened, it is in the literature, and pretending otherwise would be dishonest.

But read what Beer actually argued, and read the laws the whole discipline rests on, and the centralizing conclusion does not survive contact with them. It is ruled out — not on political grounds, but on grounds of arithmetic.

The arithmetic

The relevant principle is old and has never been refuted: only variety can absorb variety. Variety here means the number of distinguishable states a thing can be in — the full range of situations it can present you with. A regulator can control a system only if it commands at least as much variety as the system it is regulating. If it commands less, the excess does not vanish. It leaks out as failure, delay, workaround, or surprise.

Now count. A refinery unit on a bad night can present its operator with hundreds of distinguishable states: this pump cavitating while that exchanger fouls while a transmitter reads suspect and the weather drops twenty degrees in an hour. The operator, standing there, holds enough variety to absorb most of it. He has the gauges, the sound, the smell, ten years of pattern recognition, and a radio.

A vice president four levels up, holding a weekly summary, does not. He cannot. There is no amount of dashboard that gives a single remote person the variety of a hundred local situations, because the summary is produced by throwing variety away. That is what a summary is for.

So when a central authority takes the decision away from the operator, it has not improved control. It has moved the decision to a regulator with less variety than the thing being regulated, which the law says cannot work.

Two honest moves, and the counterfeit

This is the point at which the argument usually gets oversimplified into decentralize everything, and it is worth not doing that, because Beer's answer is more precise and more useful. A center does not need variety parity with its periphery. It has two legitimate instruments for closing the gap, and both of them are real engineering.

It can attenuate — deliberately reduce the variety arriving from below, so that what reaches the center is the small set of things the center can actually act on. Exception reporting is attenuation. So is a well-drawn tolerance band, a classification scheme, an escalation threshold with a number in it. Done honestly, attenuation is how a center stays sane, and refusing to attenuate is its own pathology: the executive who wants every detail and therefore understands none of them.

And it can amplify — increase the variety available at the periphery, so that more situations are absorbed where they occur. Training is amplification. So is genuine delegated authority, a decision rule the operator can apply without a phone call, a standard that encodes hard-won judgment rather than replacing it, and competent people hired and then trusted. Amplification is the move organizations underuse, because it costs money up front and pays back somewhere a budget cannot see.

The counterfeit is attenuation applied to the wrong end for the wrong reason: reducing the variety of the work rather than of the reporting, so the center can cope. Standardize the task, forbid the exception, require approval for the unusual case. It looks identical to the honest move on an org chart and it is the opposite of it, because the variety being destroyed is the variety that was absorbing real disturbances. The organization becomes governable on paper by becoming blind in practice. It is the same maneuver as closing your eyes to steady the picture.

The test is which side of the boundary the filtering happens on. Attenuate what travels upward and you have built a center that can think. Attenuate what happens at the face of the work and you have built a center that will be surprised.

A metasystem that absorbs the work of its parts has not governed them. It has eaten them, and it dies of the indigestion.

This is why centralization appears in the governance literature as a pathology. Not as a political error — as a named, recurring, diagnosable dysfunction with a predictable progression: overload at the center, starvation at the edge, decision latency rising every quarter, and a growing population of competent people who have learned that the official channel is the slow way to get anything done.

What the theory actually prescribes

The prescription that falls out is close to the opposite of the stereotype. Push decision authority down to the lowest level that holds the relevant information, because that is the only level with the variety to handle it. Give the productive units genuine autonomy inside stated bounds. Then — and this is the part people skip — build the coordination that keeps those autonomous units from colliding, and the monitoring that tells the center when a unit is drifting outside the bounds.

Autonomy without coordination is not freedom; it is five groups independently optimizing themselves into a collision. Coordination without autonomy is the approval-layer disease. The governance question is never centralize or decentralize. It is which functions require a view of the whole, and which require a view of the particular — and then placing each one where the information actually is.

A center that does this well ends up doing less than it did before, and different work: identity, direction, the long horizon, the scan for what will invalidate the current plan, and the audit that verifies the periphery is still inside its bounds. That is a demanding job. It is not the job of approving things.

If you are in the middle

Most people reading this have no authority to redesign a governance structure, and no appetite for a reorganization even if they did. That turns out to be less limiting than it sounds, because the leverage is not in the authority. It is in the information.

Almost every governance failure we have examined was carried by a broken information path before it was carried by a bad decision. The number that arrived too late. The commitment that lived in a hallway conversation instead of a document. The audit function nobody performs, so the center believes the status report. The lesson learned, filed, and never converted into a design change. You can repair any of those from the middle, without a mandate and usually without a meeting. Shorten a feedback path, write down an interface that was only ever assumed, verify one number independently — and the organization gets measurably better at knowing what is happening, which is the whole substance of control.

No committee required. That is a systems argument, and it points down and outward, not up and inward.

The variety law is Ashby's, from An Introduction to Cybernetics (1956). His own phrasing was that only variety can destroy variety; the governance literature says absorb, and means the same thing. Attenuation and amplification as the regulator's two instruments are Beer's, from The Heart of Enterprise. Centralization as a catalogued governance pathology follows Keating and Katina. The plant manager is a composite, and he is not wrong.

← Todos los apuntes