The project came up fourth on the agenda, the way it had come up fourth on the agenda for three quarters running. Everyone in the room knew the number. It was behind, it was over budget, and the thing it was built to produce had quietly stopped being the thing the company needed. Nobody said any of that. What someone said instead was that we were too deep in to walk away now. Heads moved. The project got another two quarters and another line of funding, and the meeting advanced to item five with the particular relief of a room that has agreed, together, not to look at something.
It was framed as resolve. Staying the course. Not losing nerve at the hard part. What had actually happened was simpler and less flattering. The amount already spent had walked into a decision about the future and cast a vote it had no right to cast, and the room had let it, because the alternative was to say out loud that the money was gone.
The money was gone either way. This is the fact the whole ritual is constructed to avoid. A clean decision about the project asks one thing: from here, with the work in the state it is actually in, does the next quarter of spend buy more than that quarter would buy anywhere else in the company? The sum already poured in is identical whether you stop or continue. It is the one quantity in the entire problem that cannot tell the two options apart. And it is the quantity doing all the talking.
The spent money gets a vote because stopping names a loser.
Ask why a number that cannot distinguish the options is allowed to decide between them, and you arrive somewhere no one wants to go. Stopping converts the word invested into the word wasted, on the ledger of the people who approved the investment. Continuing keeps the loss unrealized, deferred, deniable for another quarter. So the choice quietly stops being about the project and becomes about whose judgment gets indicted tonight. And the person usually asked to kill the project is the person who launched it. You are not requesting an analysis. You are requesting a confession, and then acting surprised when it does not arrive.
A whole vocabulary grows up to make the refusal respectable. Commitment. Discipline. Seeing it through. Not being reactive. Each is a genuine virtue, borrowed to dress an inability to absorb a loss in public. The organization that puts data-driven on its walls runs its single largest losing bet on a feeling it would never say in those words, which is that quitting would be embarrassing and continuing can still be called character.
What makes it almost impossible to interrupt is the arithmetic of the next step. Each new increment is small set against the total already in the ground. It is only one more quarter. It is a rounding error next to what we have committed. So the marginal ask always looks modest beside the sunk mountain, and the project is never extended in one indefensible leap. It is extended in a sequence of individually reasonable steps, each one enlarging the pile that makes the following step feel compulsory. The deeper in you are, the cheaper the next increment looks and the more expensive stopping feels. That is not a decision being made repeatedly. That is a ratchet, and a ratchet only turns one way.
Money already spent cannot be recovered by spending more of it. It can only be joined.
Here is the inversion the room never performs. Stopping does not waste what was spent. The spending already did that, on the day the project stopped being worth finishing, and that waste is a completed fact about the past. The only thing any decision can still touch is the next increment. Continuing does not redeem a single unit of the old loss. It adds a fresh one on top and calls the addition respect for the first. Every quarter the thing is extended out of loyalty to what it has already cost, the amount it has cost grows, and the loyalty that amount commands grows with it. The room believes it is protecting the investment. It is feeding it.
Then comes the quietest cost. When the project is finally stopped, almost always by someone new enough to have no signature on it, the post-mortem rarely finds the real defect. It finds scope creep, a weak vendor, a market that moved. It does not find the three consecutive reviews in which the room had the number in plain view and chose the comfortable sentence over the honest one. To write that down would be to admit the reviews were theatre, so the loss is filed as something nobody could have halted, when halting it is the precise thing a review exists to do. The organization keeps the ritual that failed and loses only the money. A review that cannot say stop is not a control. It is a receipt.
Make the next increment earn its own vote.
Rebuilding this has nothing to do with a culture of killing things fast, or with treating every hard stretch as a reason to bail. It rests on one surgical separation. The money already spent is evidence about the past. The decision is about the future. The two can sit in the same room, but they must never be allowed to answer the same question. A forward decision is made as if you were meeting the project today, at its real cost-to-finish and its real odds, with no history behind it and no author sitting at the table. Three questions carry most of the weight:
- If we were choosing to start this today, at its current cost-to-finish and current odds, with nothing already spent, would we fund it?
- Who in this room would have to concede that an earlier call was wrong for us to stop, and are they the one answering the question?
- Stated so a stranger could check it, what is the next increment actually buying, held completely apart from everything the project has already consumed?
None of this comes naturally, and it comes least naturally where it matters most. The grip of a sunk cost is strongest under time pressure, when a committed course begins to fail in public and a reputation is tied to the mast, and that is exactly the territory no slide deck and no stage-gate checklist can reach. It is the reason the organizational crisis simulation is built the way it is. Put a leader inside a scenario where a course they have already committed to starts coming apart, start the clock, and watch whether they decide forward from where things now stand or spend their moves defending the ledger behind them. The pattern is strikingly consistent, and almost entirely invisible to the person producing it, until they watch themselves produce it with the outcome still unwritten.
A project is never too big to stop. It only ever becomes too expensive to admit.