The COO is the strangest seat in the C-suite: the job description is different at every company, the border with the CEO is drawn in pencil, and success means the company runs so smoothly nobody can point at what you did. That ambiguity is exactly why new COOs fail in structural ways — the traps are built into the role before the person even arrives. The deep dive, following on from the hub article:
The early mistakes (first 90 days)
- Taking the job without a border treaty. The single biggest predictor of COO failure is skipping the awkward conversation: which decisions are yours, which are the CEO’s, and what happens when we disagree? An unwritten split guarantees turf creep in one direction or the other — and in this pairing, the collision is the most expensive one in the building.
- Promising speed before mapping the machine. Committing to delivery improvements before you know where the real bottleneck is means optimizing the visible part of the system — which is usually not the constraint.
- Mistaking motion for diagnosis. New COOs love a listening tour of the departments; fewer walk the actual flow of a customer order, ticket, or shipment end to end. The org chart lies; the flow doesn’t.
- Keeping every inherited meeting. The COO calendar arrives pre-filled with every operational review in the company. Attend them all and you’ve ratified the old operating rhythm before deciding if it works.
- Quietly outshining the CEO. COOs hired for operational grip sometimes demonstrate it by visibly out-executing the CEO in front of the leadership team. It feels like value; it reads as a challenge. The COO’s wins need to be the system’s wins.
The big failure modes
The human bottleneck
The signature COO failure, worth the full treatment. Cross-department disputes are the COO’s natural food — and eating every one personally feels like exactly what you were hired for. Six months in, there’s a queue outside your (virtual) door: every priority conflict, resource clash, and escalation waits for you to arbitrate. You’ve become the company’s throughput ceiling — and trained everyone below you that conflicts get escalated, not resolved.
The way out is mechanical, not heroic: published decision rules (what gets decided where), explicit escalation paths and RACI, and a personal discipline of answering many escalations with “what did the two of you propose?” A COO’s output isn’t decisions made; it’s decisions not needing them.
Fixing symptoms instead of systems
The launch slipped, so you personally rescue the launch. The next one slips too, because the thing that made the first one slip — unclear intake, overloaded shared team, no capacity planning — is still there. Rescue work is visible and satisfying; systems work is invisible and compounding. New COOs under pressure to show early wins consistently over-invest in rescues. The tell: the same fire, twice. One recurrence is noise; two is process.
Metric worship
Operations runs on numbers, and the failure mode is loving them wrong: pick a metric, drive it hard, and watch teams optimize the number instead of the outcome. Ship-on-time hits 99% because teams pad their estimates; support resolution time drops because tickets get closed and reopened. Every operational metric is gameable — which is why the pairs discipline from KPI examples matters doubly here (speed with quality, cost with service), and why targets need the KPIs-vs-OKRs distinction: health metrics you monitor, improvement goals you campaign on. Chasing every health metric as a goal is how you get a fast, cheap, broken machine.
Scope creep, operational edition
The COO’s remit grows the way projects balloon: every orphaned function gets handed to ops “for now” — facilities, procurement, that analytics team nobody claimed, the integration from the last acquisition. Each addition is reasonable; the sum is a COO spending Monday on real estate and Friday on data governance, with the core delivery machine unattended. Strong COOs treat their own portfolio with change control: additions are explicit trades, not silent absorption.
Shadow CEO
The darkest failure mode, and the one that ends partnerships: the COO who starts deciding things the CEO believes are strategic — entering a market “operationally,” reshaping the org “for efficiency,” pre-agreeing positions with the executive team before the CEO hears them. Sometimes it’s ambition; more often it’s drift that nobody named early (see: border treaty, above). Either way, a CEO-COO pairing survives on exactly one thing: the CEO never being surprised by which decisions got made.
What it looks like in the wild
Tim Cook as Apple’s operations chief (1998 onward) is the systems-over-rescues masterclass. Arriving at a near-bankrupt Apple, Cook didn’t heroically expedite late shipments — he rebuilt the machine: closed factories, moved to contract manufacturing, and cut inventory from months of supply to days (his line: inventory is “fundamentally evil”). The operation ran so well it didn’t need him in every decision — which is precisely the meta-KPI this article ends on, and a large part of why the board later trusted him with the CEO seat.
Sheryl Sandberg and Mark Zuckerberg (2008–2022) ran the most famous border treaty in modern business. The split was explicit from the start: he owned product and engineering; she owned advertising, operations, and policy. Whatever else one thinks of the company’s history, the structure held for fourteen years — because the border was drawn in words, up front, rather than discovered through collisions. Most CEO–COO pairings fail faster, and almost always for the want of that one conversation.
How to measure a COO: the KPIs
COO metrics are flow metrics — is the machine turning inputs into delivered outcomes, faster and cleaner over time?
| KPI | What it tells you | Healthy sign |
|---|---|---|
| On-time delivery rate | Does the company keep its promises? | High and honest — dates aren’t padded to hit it |
| Cycle time (order/idea → delivered) | How fast does work flow end to end? | Falling over time, measured on the whole flow, not one department |
| Throughput | How much does the machine actually produce? | Rising without proportional headcount growth |
| First-pass quality / defect rate | Is speed real or borrowed from quality? | Improving alongside cycle time, not trading against it |
| Cost per unit delivered | Is the machine getting more efficient? | Trending down with volume; scale actually scaling |
| Capacity utilization & bottleneck health | Where’s the constraint, and is it managed? | Known constraint, deliberately loaded — not discovered by surprise (bottlenecks module) |
| Escalations reaching the COO | Is the operating system maturing? | Falling month over month — the meta-KPI that says the machine runs without you |
That last row is the one most scorecards miss and the one that matters most: a COO whose personal decision volume rises over time is building a dependency, not an operating system.
How to use it
- “Every cross-team conflict lands on her desk — she’s become the bottleneck she was hired to remove.” (the human bottleneck)
- “Same fire, second quarter in a row. That’s not bad luck, that’s a missing process.” (symptoms vs systems)
- “They agreed the CEO/COO decision split in writing during week one.” (the border treaty)
- “Do the Zuckerberg–Sandberg split: write down who owns what before the first collision, not after.” (borders in words)
- “Fewer escalations reached her desk this quarter than last — the machine is maturing.” (the meta-KPI)
- “He answered the escalation with ‘what did the two of you propose?’ — third time this month. It’s working.” (training resolution, not escalation)
Related reading: The Biggest Mistakes New Executives Make · CEO vs President · KPI Examples That Actually Work · Scope Creep