The Sunk Cost Fallacy in Services

Every services firm has an engagement nobody will kill. It loses money in a way everyone privately acknowledges, and it survives every planning cycle, because the firm has put so much into it. That sentence, we have put so much into it, is the sunk cost fallacy in its natural habitat: the past, dressed up as an argument about the future.

The fallacy is treating money, time and effort already spent as a reason to spend more. The arithmetic against it is unforgiving. Whatever was spent is spent; it cannot be recovered by continuing, and it is not increased by stopping. The past costs exactly the same under every option on the table, which is precisely why it can never be a reason to prefer one of them.

Sunk costs are facts about the past

A sunk cost is a fact, and facts about the past belong in the accounts. Sunk costs are facts about the past; a decision is a choice about the future, and only future costs and future returns belong inside it. The moment a past number appears as an argument, the decision has been contaminated: it is no longer choosing between futures, it is protecting a feeling about history.

Why does something so obviously true get ignored so reliably? Because past spend does not present itself as a number. It presents itself as meaning: effort that would be wasted, loyalty that would be betrayed, a story that would end badly. Stopping makes the loss feel real and attributable; continuing keeps it comfortably ambiguous. This is also why the fallacy so often matures into escalation of commitment, its organisational sibling, where reputation and identity join the arithmetic on the wrong side.

Where it hides in a services firm

The fallacy has favourite hiding places in a people business.

• The engagement nobody will kill. Years of relationship investment, an alumni network of people who worked on it, a rate card quietly eroded by goodwill. The history is real. It is also not a reason.
• The practice area kept alive by its history. The firm was built on this work, so the firm keeps doing this work, long after demand has moved. Heritage deserves respect. It does not deserve a vote.
• The pursuit that must close. The bid has already consumed so many partner hours that walking away feels unaffordable, so more hours follow the first ones in.
• The internal tool that cost too much to abandon. Which is to say: it cost a lot, therefore it must keep costing.

Honouring commitments is not honouring costs

The strongest objection deserves a straight answer: we honour our commitments here. Good. A commitment is a promise about the future, and breaking it has future costs: contractual penalties, a client’s trust, the credibility of the firm’s word, the morale of the team that made the promise. All of those belong in the decision, at full weight. Honouring commitments is not the fallacy. Honouring costs is: continuing not because of what breaking the promise would cause next, but because stopping would make the past spend feel wasted.

The test that separates them fits on one sheet of paper. Two columns, both starting today: what continuing buys, and what stopping costs. Live obligations, redeployment of people, reputational consequences: all admissible, all about the future. History is not admissible. If the only entries on the continue side are what has already been spent, the decision has already been made, and the firm just has not said it out loud.

One concrete example

Clearly illustrative, with no customer implied. A firm of a few hundred people still runs the practice area it was founded on. Demand has moved to adjacent work; the practice runs sub-scale, and senior people spend unbillable weeks defending it in planning cycles with arguments that all point backwards: what it meant, what it built, who came up through it. The forward analysis takes an afternoon. Real future items: two live client contracts that run to year end and will be honoured in full; several consultants whose skills map directly onto the growing adjacent practice; a brand association that argues for a considered transition rather than an abrupt exit. Historic items: the founding story, the invested years, the memories. The decision that follows honours every commitment: the contracts are delivered, the people are retrained and redeployed, the practice is wound down over two quarters with its clients told the truth early. What was finally refused a vote was the cost.

Sunk cost in the ONX vocabulary

Decision intelligence enforces the distinction structurally rather than by willpower. In a decision run, options are priced from the present facts forward: a Scenario Run evaluates candidate options against the facts as they stand, and there is no field in it for what was spent before it ran. Claims that smuggle the past back in, such as this account is strategic, carry an evidence state: if the claim is stated rather than measured, it is visible as exactly that, and it is weighted accordingly. The past is kept honestly, as versioned facts and immutable runs, precisely so the decision does not have to carry it. That is the quiet discipline of Enterprise Decision Intelligence: the past preserved as evidence, and refused as an argument.