What is Escalation of Commitment?

Nobody decides to throw good money after bad. They decide, quarter after quarter, that this quarter is different: the recovery plan is credible now, the hard part is behind us, stopping here would waste everything the programme has learnt. Escalation of commitment is the name for that pattern: the tendency to keep investing in a failing course of action precisely because so much has already been invested in it.

The pattern has been studied in organisational research for decades, and it survives every generation of management method because it is not a knowledge problem. Everyone in the room can recite the lesson about sunk costs. They escalate anyway, because the forces doing the work are not the ones the lesson addresses.

Identity, not arithmetic

The textbook explanation is the sunk cost fallacy: spend that belongs to the past leaks into decisions about the future. In executive life a stronger force does most of the work: identity. The programme has a champion, and the champion made a public call. Stopping does not just close a budget line; it announces that the original judgement was wrong, in front of the people who watched it being made. What is sunk is not only the money. It is the reputation.

This is why escalation worsens with seniority and with visibility, and why the strongest predictor of it is structural: the person who made the original call still owns the review of it. A steering group chaired by the sponsor is not a review; it is a defence with an agenda. Asking authors to referee their own decisions asks them to convict their own judgement, and almost nobody does.

There is organisational gravity too. A programme accumulates a team, a steering committee, a line in the board pack. Killing it deletes part of the organisation chart, so the organisation defends it the way a body defends tissue. None of the people involved need to be acting in bad faith. The structure escalates on their behalf.

Where services firms escalate

In a services business the pattern has familiar shapes. The loss-making account kept for strategic reasons that nobody can state when asked directly. The troubled programme whose recovery plan always needs one more quarter. The internal platform that has consumed several roadmaps and is always nearly done. The practice bet that keeps absorbing partner time because the firm announced it loudly. In every case the increments are individually small against what stopping would appear to cost, which is exactly how the total gets large.

The ritual that breaks it

The antidote to escalation is not a tougher review; it is a review designed before the commitment, judged against pre-agreed evidence, by people who did not make the original call. Tougher meetings after the fact only harden the defence. The ritual has three parts, and all three are agreed at the moment of commitment, while everyone is still impartial about a failure that has not happened yet.

• Pre-agreed review points. The dates are written into the commitment itself, so the re-examination arrives by calendar, not by courage. Nobody has to be the person who called the meeting.
• Failure defined in advance. The commitment states what evidence would count against it: the test it must pass, the signal that would mean it is not working. A commitment that cannot fail on any evidence is not a plan; it is a belief.
• Fresh eyes in the chair. Someone without authorship runs the review. Their question is not how do we save this. It is: would we start this today, on today’s evidence?

That last question is the hinge. It reframes continuation as a fresh decision about the future rather than a defence of the past. Continuing is allowed. It simply has to win on current evidence, in front of someone whose reputation is not in the room.

One concrete example

Clearly illustrative, with no customer implied. A firm commits to building a shared delivery platform intended to serve three accounts. Two quarters in, it is late, and the steering group, chaired by the sponsoring partner, approves an extended budget because the team is close. Two quarters later it does so again. Now replay it with the ritual. At commitment, the firm wrote down two review points and the evidence that would count as failure: a named integration working end to end, and a second account signed up. At the first review, a partner from a different practice chairs. Neither test is met. The question is not whether the team worked hard; it did. The question is whether the firm would start this today, and it would not, at this scope. The programme is cut back to serving one account well, and the decision is recorded with the reasoning attached. Nobody is shamed. The original call was defensible on the original evidence. The firm simply refuses to let the past cast the deciding vote.

Escalation in the ONX vocabulary

The ritual maps cleanly onto the decision-intelligence spine. Pre-agreed review points are decision windows declared in advance. The evidence that would count as failure is written as facts with an evidence state, so at review time the claim that we are close has to survive contact with what is measured rather than what is stated. Because scenario runs are immutable, what was believed at commitment is preserved exactly, safe from hindsight and from memory’s editing. And scored outcomes accumulate in an Outcomes Ledger, so the firm learns which kinds of commitment it tends to over-defend. That is decision intelligence applied to the oldest trap in management: making the continuation of a commitment a decision, rather than a default.