Strategy · June 2026 · 8 min read
A bad number is not a decision.
When a strategic project underperforms, the financial number can’t tell you whether to cut it, fund it, double down, or shift its goal. Only one thing can.
“Some of our strategic bets are underperforming. It’s hard to know how, or when, to push back.”
A senior executive put this to me recently. I have not stopped thinking about it since.
It is the right question, asked with the wrong instrument. When I asked whether AI was the driver, the answer was honest: not directly. AI had raised the stakes, but the deeper problem was older than that.
A few of these bets had become hard to challenge at all, funded on conviction, defended by seniority, with no agreed line that would have told anyone they were off track. Most leaders answer from the financial number. But underperformance is not information. The same red number sits under a bet you should kill, one you should keep funding, one you should pour more into, and one that has already done its job.
So the decision cannot come from the number. It comes from a gate you set before the bet began: a falsifiable metric and a date that say, in advance, what result would tell you to stop. That is the answer. The rest is how to set it, and how it holds up across a portfolio.
The AI era only sharpens the need. There are more strategic bets now, their payoffs are harder to measure, and they move faster than the annual plan can track. That is precisely when a pre-committed gate matters most.
When the payoff itself is hard to quantify, you gate on the leading indicator you can name in advance, not the fuzzy end-state, and you still set the date. Most of the last two years’ bets were made on conviction and a deck, with no gate at all. Leaders are retrofitting one now. Late, and the right instinct.
I have made this call from every side of the table: selling the bet, judging the pitch, and owning the number when it came up short. The debate is always loud and the conviction high, and almost no one in the room can say what result, by when, would change their mind.
That missing sentence is the whole game.
Conviction and sunk cost look identical
Live through it, and a disciplined bet about to break out looks exactly like a doomed project nobody will kill. Years of losses, a wall of skeptics, a leader insisting it will work. You cannot tell them apart by looking, and neither can the board.
The market can’t either, until the outcome lands and everyone agrees, in hindsight, that the winners were visionaries and the losers were fools. That is survivorship bias, and it is useless in the moment. The only thing that resolves it is the gate you set in advance: with it, the bad number either cleared the bar or it did not.
Without it, you are reading tea leaves and calling it conviction.
Consultants have a name for the front half of this: the “what you need to believe”analysis. Before you commit, you write down what would have to be true for the bet to win, in numbers, by when. If you can’t, you don’t have a strategic bet. You have a hope.
The discipline is in what you do next. As the bill grows, you go back to that list and re-test it: are the beliefs still holding, or has the data quietly disproved one? Re-testing your own assumptions as you spend is what separates conviction from stubbornness. The sunk-cost operator never reopens the list.
Why the list stays closed is worth naming. A startup reopens it because the market forces the issue: miss the thesis, the money runs out, and you adapt or die. Inside a large company there is no such forcing function. Patient capital funds a falsified thesis for years, and political capital turns it into a sacred cow, a bet no one will challenge long after the numbers stopped supporting it. Failure rarely gets booked as a learning when a career is attached to it.
So the discipline a startup gets for free has to be built on purpose, and the gate is how you build it. Agreed out loud before the bet begins, it manufactures the reckoning the market would otherwise impose. It also changes what a hard call is about. When the gate is missed, retiring the bet is no longer a referendum on the sponsor’s judgment. It is their own plan working as designed.
A gate lets a board challenge a bet without it becoming a fight over who championed it.
Exhibit
Four moves, one discriminator
The same weak number; what a pre-committed gate tells you to do with it.
Without a pre-committed gate you cannot reliably tell these four apart, and the default becomes the most expensive option of all: drift.
Here is each move in the wild, and the gate underneath it.
Fire Phone · 2014–2015
The phone-as-shopping-funnel thesis didn’t survive contact with behavior. A $170 million write-down, discontinued in about a year.
The gate was missed. They stopped.
Azure, under Nadella
For years the cloud bet looked like the conviction that sinks companies. But the adoption curves were clearing the gate, so they kept funding it through the losses.
On track, unproven. They held the line.
Accelerated computing → AI
A long, patient bet on GPU computing. When AI broke out, they reallocated the whole company toward it rather than hedging.
The gate cleared early. They poured in.
DVD → streaming · closed 2023
Built to be superseded: named Netflix, not DVDflix, ran DVD as the booster rocket, then retired it after 25 years.
The objective shifted. It did not stray.
The clean cases make this look easy. The honest one does not. Meta has poured more than $80 billion into Reality Labs since 2020, a recent quarter burned roughly $4 billion on about $400 million of revenue, and the stock is at all-time highs. Discipline or stubbornness? With no falsifiable gate, no one can tell, not the board, not the market, not Meta itself, until the outcome lands. That is not a flaw in the argument. It is the argument.
The harder question: the portfolio, not the project
Which makes “do we keep backing this one project” the wrong altitude. The real question is how you run all of your bets together, and the map most of us used no longer helps. For a decade I framed portfolios in front of boards with McKinsey’s Three Horizons: defend the core, grow the adjacent, seed the future. In an AI-paced market that time axis has collapsed. A horizon-three bet can ship as fast as the core, and a rival’s moonshot can land before your three-year plan clears its first review, a point Steve Blank made back in 2019. You no longer sort bets by when they will mature, but by what would prove them wrong, and by when.
The gate is what replaces the horizon.
How you run the whole set matters more than any single call, and the evidence is unusually clear.
Read together, those three numbers are one instruction: the status quo rarely wins, the few who break out reallocate boldly rather than evenly, and the bets that survive are kept clear of the core. In practice:
- Reallocate asymmetrically. Feed the few bets that are breaking out; starve the rest. It is the highest-value move on the board, and most never make it.
- Fund in stages, against gates, not five-year forecasts. Release money as assumptions clear and redirect as they don’t, what Rita McGrath calls discovery-driven planning.
- Protect the bets from your own core.The process that makes an incumbent successful starves its disruptive bets by default, Clayton Christensen’s enduring finding. The fix is structural: separate the explorers, integrate only at the top.
Held that way, a portfolio is just a set of gated bets, sustained, accelerated, shifted, or cut as their gates clear or fail. The right number of them scales with the capital you will actually reallocate.
The mechanics are the strategy
What separates the companies that compound from the ones that thrash is not vision, and it is not nerve. It is the mechanics: the gate set in advance, the honest review when the number lands, and the willingness to cut, sustain, accelerate, or shift on evidence rather than ego.
The process is the strategy. Everything else is just conviction with a budget.
So, back to the project on your board’s agenda. You do not need to know whether it is making money. You need to know what you said, before you started, would tell you to stop, and whether you had the nerve to write it down.
A position firmly stated and loosely held. Tell me where I have it wrong.
If this is live for you
If you are sitting with a bet you cannot read, whether to cut it, keep funding it, double down, or shift it, that is the conversation I like most. Send me the bet and I will think it through with you, outside-in: a thought-starter, not a judgment.