How the best COOs think about operations

The best COOs we study share one habit: they spend more time designing how decisions get made than making the decisions themselves. By year two, the company runs without them. That is the bar.

Every operator who rises from functional head to true Chief Operating Officer makes the same pivot. They stop being the person who knows everything and become the person who designed the system that lets other people know what they need to know. The job changes from solving problems to installing the conditions under which problems solve themselves.

What follows is a distillation of the mental models that recur across Claire Hughes Johnson at Stripe, Sheryl Sandberg's era at Facebook, Keith Rabois, Harley Finkelstein at Shopify, Andrew Grove at Intel, and the texts these operators actually read.

What do top COOs actually do all day?

Top COOs spend most of their time on meta-work rather than work. They design operating cadences, write decision frameworks, clarify ownership, and remove the ambiguity that would otherwise route every judgment call back through them. A typical week is weighted toward building systems that make future weeks shorter, not toward clearing the current week's queue.

Andrew Grove's High Output Management frames this as managerial leverage. A manager's output equals the output of the people they influence, so the highest-leverage activities compound across a team. A COO who spends Monday writing a one-page decision framework that ten directors use every week has multiplied that hour by roughly 400 hours of downstream decision time across the quarter. A COO who spends Monday personally approving six decisions has multiplied that hour by zero.

Claire Hughes Johnson makes the same point in Scaling People, written after her tenure as Stripe's Chief Operating Officer from 2014 to 2021. She argues that operators at scale are in the business of installing "operating systems," by which she means the recurring rituals, documents, and decision rights that let a growing company act coherently without requiring the founder or COO to be in every room. Harvard Business Review's 2006 profile of the COO role found that high-performing COOs described their primary output as clarity. Velocity was the downstream effect.

Why do the best COOs build systems instead of relying on heroes?

The best COOs build systems because heroes do not scale, and eventually heroes quit. A company that relies on two or three exceptional people to close the books and keep customer escalations moving is one resignation away from a two-quarter recovery. Systems outlast the people who ran them last year.

Keith Rabois, former COO of Square, argues in his teaching at Stanford's Graduate School of Business that the single most important question a senior operator can ask is "what would have to be true for this process to still work if the person running it left tomorrow?" If the answer involves recruiting another hero, the system is fragile. If the answer is documentation and clear inputs, the system is durable.

Ben Horowitz covers the same territory in The Hard Thing About Hard Things. His chapter on management debt argues that most startup operational breakage is the interest payment on short-term hero dependencies that were never paid down. A founder lets one person own the commission math. Eighteen months later that person is the bottleneck on every compensation change and every deal exception. The debt compounds faster than most teams realize.

In practice, top COOs insist on written documentation of recurring processes, resist single-person dependencies on critical functions, and treat every "only Sarah knows how to do this" comment as a planning item rather than a compliment.

How do great COOs think about risk and asymmetric bets?

Great COOs think in asymmetric bets: decisions where the cost of being wrong is bounded and the upside of being right is not. They take these bets aggressively. Decisions with unbounded downside and bounded upside get deferred, decomposed, or killed regardless of how exciting they sound.

Jeff Bezos's 1997 shareholder letter drew the distinction between Type 1 decisions (irreversible, high-stakes, require deliberation) and Type 2 decisions (reversible, low-stakes, should be made quickly by someone close to the work). A common failure mode, Bezos warned, is applying Type 1 deliberation to Type 2 decisions, which produces slow, risk-averse companies that lose to faster competitors. Top COOs match the process to the type.

Matt Mochary's The Great CEO Within turns this into a practical rule: for reversible decisions, the disagreeing team should commit to the faster path and review in two weeks. The cost of being wrong for two weeks is almost always lower than the cost of spending two weeks debating. Harley Finkelstein, President of Shopify, has described Shopify's operating culture as one that budgets explicitly for reversible failure. Losing $50,000 on a reversible experiment is a feature of a healthy company, not a bug.

How do COOs make mistakes cheap to reverse?

COOs make mistakes cheap to reverse by lowering the cost of undoing a decision before they make it. They insist on pilots before rollouts, time-boxed trials before permanent hires, and reversible contract terms before multi-year commitments. The work happens upstream of the decision, not downstream of the failure.

A few tactics recur. Pilot-first rollouts: new tools go through a two to four week pilot with a single team, with rollback criteria written before launch. Time-boxed hiring for senior roles: contract-to-hire arrangements and 90-day "top grading" reviews (from The Great CEO Within) make executive mistakes recoverable. Reversible vendor commitments: annual contracts with 30-day outs, plus parallel evaluations that can be extended rather than collapsed prematurely. Verne Harnish's Scaling Up calls this "keeping optionality cheap." Explicit decommission plans: every new process gets a kill criterion written on day one so zombie initiatives do not accumulate.

What's the COO's real job in the first 90 days?

The real job in the first 90 days is listening, then mapping, then writing. The new COO is the most objective person in the building and has the shortest window to see what everyone else has normalized. Jumping to action in week three burns that advantage. Most operators we respect spend the first 60 to 75 days almost entirely on observation.

Listening means structured one-on-ones with every direct report and at least two levels below. Hughes Johnson documents a template in Scaling People: what is working, what is broken, what is the one thing you would change, who should I be talking to that I haven't met yet. Those four questions asked 40 to 60 times produce a map of the real organization no org chart captures.

Mapping is where friction patterns get written down. Weeks four through eight go to documenting how the company actually operates: how revenue is generated, how commitments are tracked, how decisions are made, where handoffs break. Teams that have lived inside the system for years have stopped seeing the friction. The new COO still can. For more on this pattern, see why teams stop noticing inefficiencies.

By day 90, the strongest COOs produce a written operating thesis: the three to five things they believe must change, why, and in what order. It becomes the reference point for the next 12 months of work. For the mechanics of sequencing that thesis, see how to build an operations roadmap.

How do COOs avoid becoming a bottleneck?

COOs avoid becoming bottlenecks by pushing decision rights as far down the organization as they can, and by building self-service systems for the questions that get asked most often. The rule of thumb we use: any question a COO answers more than twice should become a document, a dashboard, or a decision right assigned to someone else.

Peter Drucker argued in The Effective Executive (1966) that executives who answer the same operational question repeatedly are failing at the job, because the repetition is evidence that the system has not been built. Grove echoes this in High Output Management: a manager's real output is the leverage they create through system design and decision-right assignment, not the decisions they personally make.

Self-service dashboards replace weekly status asks. Written decision frameworks replace "can you take a look at this?" messages. Office hours replace ad hoc interruptions. Sandberg's Facebook-era operating style, documented in Harvard Business Review profiles, made the same point bluntly: the executive whose calendar is 90% inbound is not operating, they are being operated on.

What's the COO's relationship to the CEO, really?

The COO's real relationship to the CEO is that of a complementary operating partner who takes responsibility for everything the CEO cannot or should not be doing. The pairing works when the two roles are explicitly divided, and it breaks when they are ambiguous. Top COOs negotiate the division of labor in writing, usually within the first 30 days, and revisit it quarterly.

Ben Horowitz argues in The Hard Thing About Hard Things that the CEO-COO pairing is the most important executive relationship in a scaling company, and the one most often botched. The common failure mode is a COO hired to "handle operations" without a written definition of what that means, which leaves the CEO free to reclaim territory whenever it feels interesting and leaves the COO accountable for outcomes they do not control.

Hughes Johnson's tenure with Patrick and John Collison at Stripe is a cited example of the pairing done well. The division was explicit: the Collisons owned product vision and external narrative, Hughes Johnson owned the operating system and execution of the current plan. It survived seven years because the edges were known.

The operating rhythm matters as much as the division. Top pairs share a weekly one-on-one with a written agenda, a monthly strategic review, and a quarterly reset on priorities. The ritual is boring. That is the point.

How do great COOs reduce meetings without reducing alignment?

Great COOs reduce meetings by replacing synchronous coordination with written documents and clear decision rights. Meetings only happen when a decision requires real-time debate, when a team needs to build relationships, or when a topic is too tangled to resolve in writing. Every other recurring meeting is a candidate for elimination.

The Amazon six-page memo, popularized by Bezos and copied by many operators since, is one version of this pattern. A meeting starts with 20 minutes of silent reading. Everyone shows up prepared because the prep is the meeting. Decisions get made faster because the context is shared rather than recited.

Hughes Johnson's Scaling People recommends a quarterly audit: list every recurring meeting, ask whether its purpose is decision, alignment, or relationship, and cancel any meeting whose purpose is unclear. Most teams find 20 to 30% of recurring meetings can be cut without losing coordination. The deeper point is that meeting load is a lagging indicator of weak operating design. A company that needs more meetings to coordinate is one whose decision rights and written operating rhythm are underbuilt. For a fuller treatment, see the difference between operational and strategic work.

How do top COOs make ownership explicit?

Top COOs make ownership explicit by insisting that every recurring decision, every metric, and every cross-functional process has a single named owner with the authority to decide. Committees, rotating ownership, and "the team owns it" are treated as failure modes. A decision with three owners has zero owners.

Verne Harnish's Scaling Up uses "Who owns this?" as a foundational operating discipline. The test is simple: for any contested outcome, there should be exactly one name, written down, with authority to decide without escalation. If the answer is a list, the work stops until the list becomes a name.

Hughes Johnson extends this with DACI (Driver, Approver, Contributors, Informed) applied to every significant decision at Stripe. The overhead of writing it down was lower than the overhead of debating who should have decided afterward.

The tell of poor ownership is "we're still working that out" applied to a process that has existed for more than six months. The tell of good ownership is decisions that get made in a day by a named person and rarely get reopened. The framework for which processes need human ownership versus which can be built into systems is covered in when to hire versus automate.

Key takeaways

The COO role rewards a specific set of mental models that recur across the operators who succeed in it. Systems get built before heroes get celebrated. Asymmetric bets get taken aggressively while irreversible bets get taken slowly. Mistakes get designed to be cheap to reverse. Meta-work gets prioritized over work. Ownership gets made explicit. Meetings get cut. The CEO-COO relationship gets defined in writing.

The books operators actually read converge on the same patterns. High Output Management (Grove, 1983) for the leverage math. The Hard Thing About Hard Things (Horowitz, 2014) for the management debt framing. Scaling People (Hughes Johnson, 2023) for the operating system playbook. The Great CEO Within (Mochary) for the decision and meeting mechanics. Scaling Up (Harnish) for ownership discipline. The Effective Executive (Drucker, 1966) for the underlying logic. Different decades, same claim: the best COOs build companies that do not need them.

The operators who clear that bar have internalized the habit of designing decisions instead of making them, and they measure themselves by how rarely they get asked. For how this mindset connects to the broader discipline, see what is operations intelligence and the operations scaling playbook.