Nobody sets out to build a nine-layer company. It happens the way weight does: a good decision at a time, each individually defensible, none of them visible on a chart anyone looks at.
A star engineer wants a title, so you give them a report. A team gets big, so you split it and put a lead on each half. A senior hire needs a landing pad, so you slot them between two existing levels. Every one of those is reasonable. The compound effect is that a question asked by the person closest to the customer now takes eleven days and four meetings to reach the person who can approve an answer.
Headcount is on every dashboard in your company. Depth is on none of them. That is the whole problem, and it is also why this is one of the cheapest audits you will ever run. You can do it this afternoon, you need no consultants, and the fix does not require firing anybody.
The 40-minute audit
Step 1: Draw the real chart, not the HR one (15 minutes)
Open a blank page. Do not open your HRIS. The system of record shows reporting lines; you want decision lines, and in most companies those diverge within about eighteen months of the last reorg.
Write your own name at the top. Underneath, list everyone who reports to you. Under each of those, everyone who reports to them. Keep going until you hit people with no reports. Where a person functionally reports to someone other than their formal manager, draw the functional line and mark the formal one with a dotted stroke — you will want that discrepancy later.
Do this from memory first, then check it against the HRIS. The gaps between your version and the official version are diagnostic all on their own. If you could not name two whole branches of your own company, that is information.
Step 2: Take three numbers (10 minutes)
Maximum depth. The longest path from you to any individual contributor. Count the person at the bottom as a layer. This is your worst-case decision distance.
Median depth. Take the depth of every individual contributor and find the middle value. Maximum depth tells you about your most remote corner. Median tells you about the company you actually run.
Micro-team count. How many managers have one or two direct reports. Count them. Then, for each, write one sentence explaining why that layer exists.
That third number is usually the shock. A manager with a single direct report is not managing; they are holding a title that was created to solve a compensation or retention problem for one specific person. It was a cheap fix at the time. It is now a permanent layer sitting on top of everyone underneath them.
Step 3: The decision-latency test (15 minutes)
Numbers describe the shape. This tells you what the shape costs.
Pick three decisions your company made in the last quarter. Choose a spread: one small and reversible (a tool purchase under a thousand dollars), one medium (a change to a customer-facing workflow), one large (a hiring or roadmap call). For each, answer four questions:
- Who first noticed this needed deciding, and on what date?
- Who ultimately made the call, and on what date?
- How many people touched it in between?
- What did each of those people add — information, judgment, or authority?
That fourth question is the whole test. A person who added information or judgment earned their place in the chain. A person who only added authority — who read the thing, agreed, and passed it upward — is a delay with a title. Count how many of those you have. That number, not the org chart, is the honest measure of your depth.
What good looks like
There is no universal right answer, but the working ranges are narrower than people expect. Under 50 people, more than three layers between you and the front line is hard to justify. Between 50 and 250, four. Between 250 and 1,000, five is defensible and six is a warning. Above that, structure genuinely gets harder and the honest goal is that median depth stays well below maximum depth — a company with a max of seven and a median of four has one awkward branch. A company with a max of seven and a median of six-and-a-half has a culture.
Micro-teams have a cleaner benchmark: zero. Not because a manager with two reports is doing anything wrong, but because the structure is almost always solving the wrong problem.
The four ways layers get made — and the fix for each
1. The title fix. Someone valuable was going to leave, so you made them a manager. The layer is compensation wearing a costume. Fix: build a real individual-contributor ladder with titles and pay bands that top out above first-line management. This is the single highest-leverage structural change most companies never make, and it prevents the problem rather than unwinding it.
2. The span fix. A manager had fourteen reports, which is genuinely too many, so you split the team and added leads. Fix: before splitting, ask whether the span is unmanageable because of the number of people or because of the number of decisions routed through that one person. Often it is the second, and the answer is delegating decision rights rather than adding a layer.
3. The landing-pad fix. You hired someone senior and needed somewhere to put them, so you inserted a level. Fix: hire senior people into a defined scope, not a defined altitude. If you cannot describe what they own without reference to who reports to them, you are not ready to make the hire.
4. The acquisition or merge fix. Two teams combined and both leads kept their roles, stacked. Fix: decide the structure before the merge closes, in writing, with names in boxes. Deferring it does not make it kinder; it makes it slower and more public.
Collapsing a layer without firing anyone
This is the part people flinch at, and it is more survivable than it looks. Removing a layer and removing a person are different operations. Most of the time you are asking a manager to go back to owning work directly, keep their compensation, and lose reports — and if your IC ladder is real, that is a lateral move rather than a demotion.
Have the conversation directly and early, before the org chart circulates. A version that works:
“I want to change what you own. Right now you manage two people and spend most of your week routing decisions between them and me. I want you owning [specific scope] directly, as an individual contributor, at the same compensation and with a title that reflects the seniority. The two people currently reporting to you would move to [name]. I am telling you first because I want your read on what breaks if we do this, and because I would rather you heard it from me than from a diagram.”
Then stop talking and let them respond. Three things usually come back: relief (more often than you expect), a concern about how it reads externally, and a specific operational objection. The first needs nothing. The second is solved with a title and an announcement you write together. The third is the valuable one — write it down, because it is frequently correct.
Where this goes wrong is when the compensation quietly does not follow, or when the announcement is vague enough that people fill the gap with the worst interpretation. Both are avoidable and both are entirely on you.
A worked example
A 90-person software company. Maximum depth: six. Median depth: five. Micro-teams: four.
The latency test showed a customer-reported bug in the billing flow taking nine days to get from the support rep who logged it to the engineering manager who could schedule it. Six people touched it. Two added information — the rep who reproduced it and the engineer who scoped it. Four added only authority.
The fix was not a reorg. It was three changes: the four micro-team managers moved to IC roles with new senior titles and unchanged pay; support was given standing authority to route reproducible customer-facing bugs directly to the relevant engineering lead; and a rule that any spend under $2,000 needs exactly one approver.
Median depth went from five to four. The same class of bug now reaches an engineer in under two days. Nobody left. Nobody was paid less. The company did not announce a restructuring, because it had not had one.
The 30-day version
- Day 1: Run the audit. Write down your three numbers where you will see them again.
- Days 2–5: Run the latency test on three real decisions. Count the authority-only links.
- Days 6–10: Write the one-sentence justification for every micro-team. Circle the ones where the honest sentence is about compensation.
- Days 11–15: Build or fix the IC ladder. You cannot collapse title-fix layers without somewhere for those people to land.
- Days 16–25: Have the conversations. One at a time, in person, before anything is written down publicly.
- Days 26–30: Publish the new structure and the new decision rights together. The second matters more than the first and gets a tenth of the attention.
- Day 90 and every quarter after: Retake the three numbers. Depth rebuilds silently, and the only defence is measuring it on a schedule.
Failure modes
Cutting depth without granting authority. If you flatten the chart but every decision still routes to you, you have not shortened the path — you have made yourself the bottleneck for more people. Layers and decision rights have to move together.
Confusing this with cost-cutting. A layer audit that becomes a headcount exercise will be read as a layoff in disguise, and you will get the fear without the speed. If you are cutting costs, say so separately and honestly.
Doing it once. Depth is not a state, it is a drift. Companies that fix this in a single dramatic reorg are usually back where they started in three years, having spent enormous political capital to arrive at the same place.
Benchmarking against companies far larger than you. A 10,000-person organisation has structural constraints you do not have. Borrowing its layer count is how a 60-person company ends up with five.
What this does not fix
Flattening does not fix bad managers, unclear strategy, or a company where nobody knows what they are optimising for. Depth is a multiplier on those problems, not their source. A shallow org with no strategy just arrives at the wrong answer faster.
But it is the rare organisational problem with a genuinely cheap diagnostic. Three numbers, forty minutes, no budget. Most companies have never taken them once.