Why control stopped working
Decentralization is usually argued as a matter of values — trust people, flatten hierarchy, be modern. That argument is weak, because someone can simply disagree. The stronger argument is structural: command-and-control was the right answer to a specific situation, and that situation is gone.
The exception that we mistook for the rule
Management as we know it — planning separated from doing, decisions rising through a hierarchy — was invented around 1900 and worked extraordinarily well. The mistake is assuming it worked because it is inherently good. It worked because of the conditions of that particular era.
Seen across the whole span, the industrial era is a dip — a temporary trough of low market dynamics. Command-and-control is the machinery built for that trough. We are now out of it, still operating the machinery.
Complicated is not the same as complex
This distinction does more work than any other idea in the field, and it is routinely blurred.
⚙️ Predictable
- Cause and effect can be traced, given enough expertise.
- The same input reliably produces the same output.
- Can be described as a rule — and therefore automated.
- A jet engine is complicated. So is payroll.
🌱 Surprising
- Cause and effect are only visible in hindsight, if at all.
- The same action produces different results on different days.
- Cannot be reduced to a rule — only met with judgement.
- A market is complex. So is a team.
Every organization contains both. The complicated parts should be standardized, documented and automated — that is what processes are good at. The complex parts cannot be, and the attempt is what produces the familiar pathology: ever more detailed rules for situations the rules never anticipated.
Gerhard Wohland's sharpest observation follows from this: under pressure, organizations reach for their processes first, because that is the familiar territory. But process is the instrument for the predictable half. Applied to the unpredictable half it does not merely fail — it inflames, producing exception rules, escalations and approval layers that make everything slower.
Three reasons control fails specifically
The information is at the edge
Whoever talks to customers knows first what is changing. In a hierarchy that knowledge has to travel up to reach a decision, and every step costs time and fidelity. By the time it arrives, the decision is late — and it is made by someone further from the facts than the person who reported them.
Targets get gamed, not met
W. Edwards Deming put it bluntly: people with targets their job depends on will meet those targets, even if they have to damage the enterprise to do it. This is not a character flaw. It is the predictable response to being measured on a number rather than a result.
The system beats the people
Deming's other estimate: roughly 94 % of problems come from the system, only about 6 % from individuals. Which means most performance management works on the smaller half — and that replacing people rarely changes the outcome, because the new person inherits the same system.
Why it persists anyway
If the case is this clear, why is almost every company still built the old way? Three honest reasons, none of them stupidity.
It works well enough. Command-and-control does not collapse; it just gets progressively more expensive. Decisions get slower, coordination costs rise, capable people leave. None of that shows up as a single crisis, which means there is rarely a moment that forces the question.
It is legally and financially legible. Auditors, banks, boards and regulators all expect an org chart, a budget and a signature hierarchy. Anything else requires explanation, and explanation costs.
The alternative is genuinely harder. Hierarchy is a solved problem you can hire for. Decentralized structure has open questions — careers, pay, accountability — that each organization has to answer for itself. That is real work, and it is reasonable to be wary of it.
What this means
The argument in this article is deliberately not moral. Nobody has to believe in flat hierarchies or trust as a virtue. The claim is narrower: a structure designed for predictable markets performs badly in unpredictable ones, and that is a question of fit rather than of conviction.
Which also means the honest version admits the reverse. Where a market really is stable and the work really is mostly complicated, the traditional model can be the right one. Dave Snowden makes exactly this point — ordered work is a legitimate domain, not a defect. The mistake is applying one answer everywhere, in either direction.
Everything else in this library follows from here: if the market is unpredictable, decisions belong where the information is. What that looks like structurally is the subject of the next chapter.
Sources
- Niels Pflaeging: Organize for Complexity. BetaCodex Network, 2012 — the historical arc of market dynamics.
- Gerhard Wohland, Matthias Wiemeyer: Denkwerkzeuge der Höchstleister. 2012 — dynamics versus complication, and why process fails on the living half.
- W. Edwards Deming: Out of the Crisis. MIT Press, 1986 — the 94/6 estimate and the critique of numerical targets.
- Frederick W. Taylor: The Principles of Scientific Management. 1911 — the original, worth reading to see how reasonable it was for its time.
- Dave Snowden, Alessandro Rancati: Managing complexity (and chaos) in times of crisis. EU JRC, 2021 — on ordered domains being legitimate.
More from the Beta Library: betaos.org/library