Every org chart is an argument about how decisions should flow through a company. Most of the time nobody states the argument out loud, which is why so many reorganizations change the picture and not the behavior.

I've been through enough of these to have a strong opinion: if you can't articulate which decisions get faster and which get better under the new structure, you haven't designed anything. You've rearranged.

Structure is a bet about coordination cost

Here's the underlying logic, stripped of the usual language.

Any company has work that requires coordination between people. Structure is how you decide which coordination happens inside a team — cheaply, informally, in a hallway — and which has to happen across teams, which is slow, formal, and expensive.

You cannot eliminate coordination cost. You can only choose where to pay it.

Organize by function and coordination inside each function is nearly free, while anything requiring several functions to move together becomes a project with meetings. Organize by product or customer and cross-functional work inside a unit gets fast, while consistency and expertise across the company become the expensive thing.

There is no structure without a cost. The question is only whether you chose which cost to pay, or inherited it by accident.

So the real design question is: what coordination is most critical to our next eighteen months, and does the structure make that coordination cheap?

The reorg that doesn't work

The failure pattern is consistent. A leadership team feels friction, concludes the structure is wrong, and redraws it. New boxes, new titles, new reporting lines. Announcement, brief disruption, and then — six months later — the same friction in a slightly different location.

That happens when the reorg addressed reporting lines without addressing three other things that actually determine behavior:

Decision rights. Who can decide what, alone, without consulting anyone? If this doesn't change, nothing changes. Reporting lines describe who someone talks to; decision rights describe what they can do. Most reorgs alter the first and leave the second untouched.

Metrics. People optimize for what they're measured on, and if the measures don't move with the structure you get teams organized one way and incentivized another. That's not a reorg, it's a contradiction with a new chart.

The meeting cadence. Structure describes the formal path; the meeting schedule describes the actual one. If the same eight people meet every Tuesday to make the same decisions they always made, the org chart is decoration.

Move those three and you can often get what you wanted without redrawing anything.

What good structure does for a growing company

At the stage I work with most — companies past product-market fit, wrestling with the operating model — a few principles hold up consistently.

Structure to the constraint. Whatever is currently capping growth should get the cleanest, most direct structure and the shortest path to the CEO. Everything else can be less optimal. Trying to make the whole chart elegant is how you end up with something balanced and slow.

One owner per outcome, always. If two people share responsibility for a number, neither has it. This is worth accepting real awkwardness to preserve.

Span of control is a symptom, not a target. A manager with fifteen reports isn't automatically overloaded — it depends entirely on how much decision-making those reports need to route through them. A manager with four reports who all escalate constantly is more overloaded than one with fifteen who don't. Fix the escalation pattern before you add a layer.

Add layers late and reluctantly. Every layer adds a translation step and a place for information to get sanded down. Some are necessary. Most companies add them earlier than they need to, usually to solve a title problem rather than a coordination problem.

The title problem, since it's usually lurking

A lot of structural decisions are actually compensation and status decisions wearing a costume.

Someone deserves recognition, the only recognition mechanism you have is a bigger title, a bigger title implies more people, so you build them a team. Now the structure reflects who needed rewarding rather than how work should flow.

The fix isn't complicated but it does require deciding in advance: build a genuine senior individual-contributor track with real money and real status attached, so that "this person is excellent and should be paid and recognized like it" doesn't have to be answered by giving them direct reports.

Companies that don't do this end up with an org chart that's an archaeological record of past retention crises.

How to know whether it worked

Six months after a structural change, ask a specific question rather than "how's it going."

"Name a decision you made in the last month that you would have had to escalate before."

If people have ready answers, the structure changed something real. If they think about it and can't come up with one, you moved boxes.

That's the whole test, and it's worth asking before the reorg too — as a prediction. Write down which decisions should move down, to whom. Then check. Most reorgs never generate a falsifiable claim of any kind, which is exactly why so many of them quietly fail without anyone concluding they did.