Hiring a COO Won't Fix the Founder Bottleneck: The Three Seats You Have to Move
You hired the COO nine months ago. The reference calls were excellent, the first ninety days went well, and the operating cadence is visibly better than it was.
And you are still the person who settles whether sales or delivery wins the argument about the March deadline. You still get the Slack message that starts with “quick sanity check”. Your calendar looks almost exactly the way it looked last year, except that you now pay a senior salary for the privilege.
Nothing went wrong in the hire. Hiring a COO is a structural move made against a structure nobody drew, and what happens by default is that the work moves and the authority does not. A title transfers a job description. It does not transfer the three seats the founder is actually sitting in.
- A COO appointment moves the work and leaves the decision rights. The new executive runs the operating cadence while the founder keeps the right to decide and the right to settle disputes, so the queue at the founder's desk does not get shorter.
- The Three Seats: every founder occupies an operator seat (running the work), a decider seat (holding the right to close a named set of decisions), and an arbiter seat (settling it when two owners disagree). A COO hire reliably moves the first. The other two move only if someone writes them down.
- The arbiter seat is the one I see stay. Two function heads who disagree keep walking to whoever settled it last time, and no announcement changes that, because they are not asking for permission, they are asking for a ruling.
- Donald Hambrick and Albert Cannella found the opposite of what contingency logic predicts: across a ten-year sample, CEOs who had COOs delivered lower organizational performance than those who did not, and no conditions emerged under which the arrangement reliably paid off.
- Jeremy Marcel, studying 153 firms in five industries, found the presence of a COO was positively related to return on assets and to market-to-book ratio, contingent on the wider top management team. Two careful studies, opposite headline results, which is what you would expect if the appointment itself is not what decides the outcome.
- What to write before day one: the decision inventory (what you currently decide, from your last month of calendar and sent mail), the seat map (which of those decisions moves, which you keep and why), and the arbitration rule (who rules on a seam dispute, and what standard they apply).
Why doesn't hiring a COO fix the founder bottleneck?
Because a bottleneck is a property of the decision system, and hiring a COO adds a person to the org chart without changing where decisions are allowed to close. The work redistributes. The authority does not, unless someone moves it deliberately.
Look at what the hire actually does on day one. Your COO takes over a set of functions, meetings and reports, all of it real and all of it useful. But the decisions that reach you do not reach you because of the reporting line. They reach you because at some point you were the only person who could settle them, everyone learned the route, and the route was never closed. A new box on the chart does not close a route. It adds a box the route can flow around.
This is the harder version of how to delegate decision-making authority. Delegating one class of decisions to one leader is legible: you can write the boundary on one page. A COO hire is a portfolio handover, dozens of decision classes at once, and the size of it is what makes people skip the writing. It feels absurd to write an inventory for something as large as “operations”, so nobody does, and the portfolio arrives as a job title with an implied scope that every person in the company fills in differently.
There is a second reason, and it is the one founders least enjoy hearing. Being the person everything routes through is uncomfortable and also legible: you know your day is full and you know why. Decisions you care about being closed by somebody else, in rooms you are not in, is a quieter state and much harder to sit with. Some of the reaching back in month four is not habit. It is a preference that was never stated out loud, and an unstated preference usually beats a written org chart.
What does the research actually say about CEOs who have COOs?
It disagrees with itself, and the disagreement is the useful part.
Donald Hambrick and Albert Cannella set out to test a sensible contingency argument: that COOs should help most where industry dynamism is high, where the organization carries extraordinary task demands, and where the CEO's own background leaves a gap. They found partial support for that logic in explaining who ends up with a COO, since CEOs lacking operational experience and CEOs new to running the firm were relatively more likely to have one. They found none at all for the same logic explaining when the arrangement worked. Instead they report a substantial negative main effect: CEOs who had COOs delivered lower organizational performance than those who did not (Hambrick & Cannella, 2004).
Five years later Jeremy Marcel looked at the same structural form through an upper-echelons lens and reached a different headline. Across 153 firms in five industries, the presence of a COO was positively related to two performance measures, return on assets and market-to-book ratio, and those relationships were contingent on the characteristics of the wider top management team (Marcel, 2009). His argument for the mechanism is worth noting: a COO can improve how thoroughly the top team processes information, partly by reducing how much power is centralized around the CEO.
Both studies look at large corporations with public financials, not at owner-managed firms of forty people in Cluj or Copenhagen, and both are observational. Neither tells you what a COO would do in your company. Read them for the shape, not the number.
The shape is this. Two careful studies of the same structural arrangement produced opposite headline results, which is hard to explain if the appointment itself determines the outcome, and easy to explain if what matters is what moves along with the title. Marcel's mechanism points that way: the benefit he describes comes from decentralizing power away from the chief executive, which is a structural change rather than a hiring event. That reading is mine and neither paper tested it. It does mean the useful question is not whether to hire a COO, but which seats you are prepared to vacate.
The Three Seats: operator, decider, arbiter
Every founder of a growing company occupies three seats at once. They feel like one job, they sit in one diary, and they come apart only when you try to hand them over.
- The operator seat. Running the work: the cadence, the reporting, the functions, the hiring plan, the fires. This is the seat the job description describes, and it is the one a COO hire reliably takes. It is also the only one most handovers ever address.
- The decider seat. Holding the right to close a named set of decisions without checking. Pricing exceptions above a threshold. Which of two roadmap commitments slips. Whether a client gets custom terms. This seat does not move with a title, because a title says what somebody manages, not what they may close. Moving it takes a written inventory and a stated limit: the Delegation Envelope applied across a portfolio rather than one class.
- The arbiter seat. Settling it when two owners with legitimate claims disagree, and saying what standard the ruling used. Sales wants the date, delivery wants the scope, both are right inside their own mandate, and somebody has to rule. In my experience this is the seat that stays put, and it keeps the founder in the middle of a company they believe they have handed over.
Why the third seat is so sticky is worth sitting with. When a function head escalates a seam dispute, they are not asking permission to do their job. They are asking for a ruling, and rulings only bind if the person making them can enforce the outcome across both functions. If the two heads believe the founder will revisit the COO's ruling when it goes against a function the founder cares about, the COO's ruling is provisional, and a provisional ruling is worth less than the trip to the founder's office. They do the arithmetic quickly and correctly, and the old route re-forms.
What follows is a recurring shape I meet across engagements rather than one company's story, offered as a pattern illustration and not as a measured outcome. An operationally strong COO arrives, the meeting cadence improves within a quarter, the founder's diary briefly empties, and by month five the escalations have returned in a new costume. They no longer arrive as “can you approve this”. They arrive as “can you weigh in”, which is the same request with better manners.
Three seats, and the first one moves for free. The other two are work.
What has to be written before the COO starts?
Three artifacts, none of them long, and all three cheaper to write before the hire than to reconstruct after it.
The decision inventory. Go through the last month of your calendar and your sent mail and list every decision you actually closed. Not your job description, the evidence. Memory is a poor instrument here, because each individual decision was small and felt like helping. What usually surprises people is the composition rather than the count. Sort the list by how often each decision recurs and by what it would cost to get one wrong.
The seat map. Against each decision class, write one of three letters. M for moves to the COO. K for kept by you, with the reason written next to it. S for shared, which should be the smallest pile, because shared means both of you and in practice means neither of you. The reason column on the K pile matters more than the letters do. “Board relationships” is a real reason. “I have always done this” is not a reason, it is a description, and it is how a K pile quietly grows until the M pile is decorative.
The arbitration rule. For each recurring seam between two functions, name who rules and what standard they apply. This is the Seam Contract applied to the handover: the rule has to name the tiebreaker and the standard, because a tiebreaker without a standard is just a different person guessing. Then say the part that costs you something: that the ruling stands, and that if you disagree with one you change the rule going forward rather than reversing the instance. A ruling you overturn once is not a ruling, and everyone watching learns the real structure that afternoon.
Then tell the company, not in an all-hands slide about the new COO's remit, which people forget by Thursday, but in the moments where the old route gets used. The first few times somebody brings you an M-pile decision, the answer is a redirect, even when you know the answer and it would take eleven seconds. Especially then. Those redirects are the transfer; everything else is paperwork.
How do you tell, six months in, whether the seats moved?
Count, rather than ask. Everyone involved will tell you the transition went well, and they will mean it.
- Pull a fresh month of your calendar and sent mail. Same method as the inventory, six months later. The impression of a successful handover survives a lot of contrary evidence, so work from the record.
- Sort what you closed into the three seats. Operator decisions you should not still be making. Decider decisions on the M pile that came to you anyway. Arbitrations between two functions that you ruled on.
- Read the arbitration pile first. It is the smallest and it carries the most information. If you are still ruling on seams between functions that both report to the COO, the arbiter seat did not move, whatever the operating cadence looks like.
- For every M-pile decision that came back, find the missing input. What did the COO need that only you had? A relationship, a piece of board context, a customer history. That answer is the next line of the handover, and it is usually an information problem rather than a confidence problem.
- Change one thing, not five. Take the decision class that cost you the most hours this month, move it properly, and count again in a month.
This is what Leadership Architecture looks at from the other direction. Escalation Discipline asks whether decisions travel upward on a defined trigger or by habit. Leadership Load Balance asks whether decision weight is distributed across the executive team or concentrated. Decision Authority Dependency, which we treat as a moderator rather than a capability, asks how much of the system still runs through one person regardless of what the chart says. A COO hire puts all three under load at once, and chronic escalation after the hire is the reading. If the pattern is broader than the handover, the frame that fits is scaling a founder-led company rather than the hire itself.
Here is the question I would ask before you open the search. Not “can this person run operations”, because you will screen well for that. Ask instead: which decisions am I prepared to be wrong about, in writing, without reaching back? If the honest answer is none of the ones that matter, the search is not the next step, and no candidate on the market will fix it.
Frequently asked questions
Why doesn't hiring a COO fix the founder bottleneck?
Because the bottleneck is a property of the decision system rather than of the org chart. A COO hire moves the work, the functions and the reporting, but decisions reach the founder along routes that formed when the founder was the only person who could close them. A new box does not close an existing route, so decisions flow around it unless someone writes down which ones move and which ones stay.
What are the three seats a founder has to hand over?
The operator seat (running the work, the cadence and the functions), the decider seat (the right to close a named set of decisions without checking) and the arbiter seat (settling disputes between two function owners who both have a legitimate claim). A COO appointment reliably moves the first. The second moves only with a written decision inventory and a stated limit. The third moves only if the founder stops reversing rulings.
Does the research show that hiring a COO improves performance?
The two best-known studies disagree. Hambrick and Cannella (2004) report that CEOs who had COOs delivered lower organizational performance than those who did not, with no conditions emerging under which the arrangement reliably paid off. Marcel (2009), across 153 firms in five industries, found the presence of a COO was positively related to return on assets and market-to-book ratio, contingent on the wider top management team. Both look at large corporations with public financials and both are observational, so neither predicts what would happen in a specific company.
What should I write down before a new COO starts?
Three things. A decision inventory built from your last month of calendar and sent mail rather than from memory. A seat map marking each decision class as moving, kept (with the reason written out) or shared. And an arbitration rule naming, for each recurring seam between two functions, who rules and what standard they apply. All three are cheaper to write before the hire than to reconstruct afterwards.
How do I know six months later whether the handover actually held?
Count rather than ask. Pull a fresh month of calendar and sent mail, sort what you personally closed into operator, decider and arbiter decisions, and read the arbitration pile first. Still ruling on disputes between two functions that both report to the COO means the arbiter seat never moved, whatever the operating cadence now looks like.
Sources
Hambrick, D. C. & Cannella, A. A. (2004), CEOs who have COOs: contingency analysis of an unexplored structural form, Strategic Management Journal, 25(10), 959-979
Marcel, J. J. (2009), Why top management team characteristics matter when employing a chief operating officer: a strategic contingency perspective, Strategic Management Journal, 30(6), 647-658
Client examples are anonymized composites; no figures are invented.
Which of the three seats are you still sitting in?
Five minutes. No account. A structural read on whether your decision system, not your team, is the real constraint.