The COO: What Are We Actually Optimising?

Hannah Wolcott reads the Vantor email at 07:04 and forwards it to Ravi Menon at 07:15 with one line: this is not recoverable inside the window. Ravi is the chief operating officer. He does the arithmetic on his phone in the car park.

Five hundred and twenty boards, eighty a week: six and a half weeks of production on the shelf, then 440 ovens that will not exist, against 960 firm orders and about $5.3m of revenue. He has an hour before the management meeting, and less before somebody in sales hears a rumour. What he does with it shapes every decision the next nine people take.

What lands on Ravi’s desk

What lands is a forwarded message and Hannah’s judgement that no rescue arrives in time. The arithmetic is fixed: Vantor is the sole source of the CB-40, stock is 520 boards, the run rate 80 a week, the delay twelve weeks, quarter end in week nine.

Nothing else is fixed: which customers wait, what Calder will spend, who decides, what anyone is told. Those are choices, and choices need something to be measured against. Ravi decides what “good” means for twelve weeks, in a sentence nine people can apply without him in the room.

What a chief operating officer actually does

A COO owns how the company delivers what it sold — factory, supply chain, planning, service. Most weeks that is steady adjustment: capacity against demand, this month’s schedule against next month’s.

A shock changes the job rather than intensifying it. There is now a question no single desk can answer, because it trades one department’s success against another’s: is Calder saving the quarter’s revenue, its cash, or its customers? Each has an owner arguing in good faith, and only the COO sits above all three.

The software on Ravi’s desk

Ravi’s tooling is deliberately thin. The dashboard exists so that eleven people argue about what to do rather than about whose figure is right — in a crisis the cost of two competing spreadsheets is measured in days. The shared channel replaces the meeting sprawl that a shock otherwise produces: one place, one time, the same small group, and a written trail for everyone who was not in the room.

The decision log is the one people skip and later wish they had not. Twelve weeks from now, when the board asks why $85,000 went on air freight, the useful answer is a dated entry naming the decision, the evidence and the alternative rejected. It is also how a crisis stops being folklore: without it, the only record of how Calder survived this is what people happen to remember.

The decisions

Naming the objective

This is the most valuable thing Ravi does all quarter, and it takes twenty minutes.

Three objectives are defensible and they conflict. Maximise revenue puts every board into the most valuable oven Calder sells, which starves the Compact line and leaves Nordfoods’ 240 ovens unbuilt — and Nordfoods moves that volume permanently the first time Calder is late. Protect cash spends nothing: no $85,000 expedite, no $180,000 of Kestrel tooling, no broker boards at $1,100. Easy to defend in a finance meeting, and expensive over a year, because it buys back no time. Protect every relationship fails on arithmetic: 520 boards cannot serve 960 orders, and spreading the pain evenly still loses Bellwether, who cancels on a partial delivery.

So Ravi writes it down, and it binds everyone downstream for twelve weeks: protect first the relationships that cannot be bought back at any price; recover revenue second; spend money only where it buys something lasting beyond the quarter.

Read it as three rulings. A customer who leaves for good outranks one who is merely angry, so Nordfoods and Bellwether outrank a larger order with a capped penalty clause. Revenue matters, and comes second. And money buys only what survives week twelve: expedite that rescues Bellwether’s dry dock qualifies, broker boards do not.

That sentence is what lets nine people decide without asking him. Tomasz can rank orders and defend the ranking; Claire can approve a spend that never pays back inside the quarter. Without it, procurement optimises for unit cost, planning for margin per board, finance for the quarter, sales for the largest account — each locally rational, all contradictory.

Who runs this

The second decision is structural and takes ten minutes. Ravi names Tomasz Nowak as owner of the plan, convenes a standing group of seven — procurement, engineering, planning, manufacturing, demand, finance, sales — and fixes a fifteen-minute checkpoint at 08:30 daily, same time, same people, changed or not. One board count, one order book, one allocation.

Each element does work. A named owner means a person whose week this is, not a topic several people are worried about. A small group can decide; twenty interested parties make a briefing. A fixed time means nobody has to call a meeting to have one, and no question waits more than a day. One set of numbers matters most, because the real disagreement in a crisis is rarely about what to do — it is about what is true.

Ravi deliberately does not chair it. He owns the objective and the escalations; Tomasz owns the plan. A COO who runs the daily meeting becomes the bottleneck, leaves nobody above the chair to escalate to, and stops running the rest of a $50m company.

What this replaces is the crisis where everyone is on it: three spreadsheets, two people ringing the same supplier, nobody sure who is telling Harlow anything, urgency high and decisions zero. Everyone being on it is indistinguishable from nobody being on it.

What not to do yet

The third decision is a list of prohibitions, harder to hold than it looks, because on day one doing something feels like leadership.

Until the arithmetic is verified and the objective circulated, nobody promises a customer a date, nobody raises a purchase order with a broker, nobody makes a public statement. The 520 is a system figure somebody must physically count, and Yusuf has not checked whether all 960 orders are firm rather than forecast dressed as fact. Committing $220,000 to broker boards before knowing whether the 520 already covers the accounts that matter is not decisiveness; it is buying uncertified electronics for a gas appliance to feel busy.

The specific danger is early, well-meant reassurance from salespeople. An account manager hears the rumour at 09:00, rings Harlow Hotels at 09:30, and says something sincere: we’re on it, worst case a week or two. By Friday the real answer is six weeks — and Harlow has re-planned a refurbishment, closures and sub-contractors included, around a date that never existed.

So the delay costs less than the promise. Three days’ wait for a real date is an irritation; an unkeepable promise is spent credibility, and Calder needs it in week ten to ask that customer to accept a phased delivery.

What he tells the chief executive

By 08:30 Ravi has spoken to Eleanor Vance, and what he gives her is four things and not a plan.

The size of the exposure: about $5.3m of revenue and $1.7m of gross profit against a $12.5m quarter, quarter end in week nine. The range of outcomes: at best most of it slips into next quarter; at worst Bellwether cancels, Nordfoods moves its volume for good, and penalties sit on top. The known unknowns: whether all 960 orders are firm, whether Vantor’s twelve weeks holds, whether any alternative source qualifies in time. And when she hears next: a full picture on Friday, a daily note in between.

The temptation is to bring a plan, because a problem with no solution feels like a loss of control. But a plan built before the facts are checked is a guess wearing a suit, and once stated it is what everyone measures reality against. There is a colder reason too: a leader who over-promises on day one loses the authority to spend money on day ten. Tell Eleanor this is mostly recoverable, and every later request — $85,000 of air freight, $180,000 of tooling — reads as an admission that Tuesday was wrong. Honest early estimates buy expensive late decisions; optimistic ones mortgage them.

What a shock reveals about an operating model

A steady month lets an organisation hide behind its structure; a shock strips that away. At Calder the person who knows the true board count is a stores supervisor who cannot spend anything, and the person who can authorise $180,000 of tooling has never seen the order book. What decides the speed of response is the distance between those two, and the incident structure exists to shorten it.

Where this goes wrong

The heroic version looks impressive for about a week. The COO chairs everything, takes the supplier calls, approves each shipment. Decisions are fast while he is in the building and stop when he is not, and there is no escalation route left because the escalation point is chairing it.

The quieter failure is worse. The objective is stated once in a long email and never used again, so by week four different desks work to different rules and each can honestly say they are following the plan. Ask four people in week six what Calder is optimising for; four answers means you never had an objective. You had a paragraph.

What Ravi hands on

The objective and the structure go to all ten desks in front of him — the nine who will act on it and the chief executive who will fund it. They arrive before anyone has anything to report, which is the point.

To Hannah goes a sourcing brief, deliberately wider than “find boards”: price and date every route, including the ones Calder will not take, and say why not, so options are killed on the record rather than left hanging. That is module three.

To Dev goes a feasibility question with a permission attached: can the oven be built with a substitute board, and how long including safety recertification — and a fast, evidenced no is a valuable answer, not a failure. That is module four.

Everyone inherits the constraint that makes the rest of the story possible: no customer hears a date until Friday. Tomasz’s allocation in module five and Marcus’s protected list in module nine both depend on nobody having spent that promise first.

Spot the decision

Read each situation and decide how a COO should handle it, then tap a card to check.