The structure is the routing table
Every reporting line is a pre-commitment about which decisions travel up and which stay down. When you draw a box and connect it to another, you are not describing communication; you are setting a default about where a given class of judgment must be ratified. The real mechanism is escalation cost. A decision crosses a boundary whenever the person closest to the information lacks the authority to act on it, and each crossing adds a queue, a translation loss, and a delay measured not in the meeting itself but in the wait for the meeting. Most organizations believe they are slow because their people are slow. They are slow because their structure routes too many decisions through too few nodes, and those nodes are congested by design.
This reframes the org chart as an allocation of decision rights, not headcount. The useful question is never "who reports to whom" but "what can this role decide alone, what can it decide after consulting, and what must it send upward." A structure is well designed when the decisions that are frequent, reversible, and locally informed are settled locally, and only the rare, expensive, irreversible ones climb.
The trade-off between speed and coherence
Pushing decision rights down buys speed and costs coherence. Pulling them up buys coherence and costs speed. There is no arrangement that escapes this; there is only the arrangement that matches the decision to the level that should own it. The governing variable is the cost of being wrong. A reversible decision made fast and corrected is almost always cheaper than the same decision made slowly and correctly, because the slow path also incurs the cost of everything that waited behind it. An irreversible decision inverts the math entirely.
So the design rule is not "decentralize" or "centralize" as a temperament. It is to sort decisions by reversibility and informational locality, then place authority accordingly:
- Frequent and reversible: delegate fully, accept variance, measure outcomes rather than approve inputs.
- Rare and irreversible: centralize, slow down deliberately, require dissent before commitment.
- Frequent but coupled: do not delegate the decision; delegate within a standard that holds the interfaces constant.
The accountability failure mode
The characteristic pathology of poor decision architecture is the diffusion of ownership across a committee. When a structure routes a decision through several roles that must all agree but none of whom can be held singly responsible, you have manufactured a body that can block but cannot decide. Consensus structures feel inclusive and are, in slow conditions, defensible. Under pressure they fail in a specific way: the decision is made by no one, the failure is owned by no one, and the lesson is learned by no one. Accountability requires a name attached to a choice. A structure that cannot produce that name after the fact did not have accountability; it had the appearance of it.
The related failure is the reorganization as avoidance. Leaders who will not confront an underperforming individual or an unresolved strategic conflict often redraw the boundaries instead, hoping the new lines will resolve what the old ones exposed. The chart changes; the unmade decision remains. Reorganizing is a legitimate tool for re-routing decisions, and an expensive way to postpone one.
The decision implication
Design the structure backward from the decisions it must produce. Begin not with the boxes but with the ten or twenty consequential choices the organization makes repeatedly, ask where the information for each one actually lives and how costly each is to reverse, and then place authority where the answer points. Treat the org chart as the artifact, not the input. Two further commitments make this real: every consequential decision should have one accountable owner even when many advise, and the right to decide should sit with the person who holds the context, not the person who holds the title. When speed and accountability both decay at once, the cause is almost never the quality of the people. It is a routing problem, and routing problems are solved by redesigning the routes.