The Commitment You Should Not Have Made
Why this matters
Most owners can name the last job they should have walked away from. Almost none can name the last commitment they should not have made, because commitments do not arrive with a quote attached and they are agreed to in under ten seconds, usually while doing something else. They are also the single largest source of hours an owner cannot account for at the end of a month.
This card is a taxonomy: the six shapes a bad commitment takes, what makes each one seductive in the moment, and the tell that is visible before you answer. Declining work that does not fit your shop is a different question with its own screen, and getting out of something you already agreed to is a procedure covered elsewhere. This is about recognising the shape while your mouth is still open.
A bad commitment is not one that went wrong
Sort this out first or the whole taxonomy collapses into hindsight. Plenty of good commitments go badly: the customer turns difficult, the supplier misses, the scope moves. That is risk, and you priced it.
A bad commitment is one that was going to cost more than it looked like at the moment you agreed, for a reason that was knowable at that moment. The test is not the outcome. The test is whether, at the instant of agreeing, you could have named three things: the finish line, the hours, and who does it if you are out for two weeks. If you could not name all three and you agreed anyway, it was a bad commitment even in the version where it went fine.
The six shapes
| Shape | What it sounds like | What it actually creates | The tell |
|---|---|---|---|
| Open-ended availability | "Call me anytime" | A standing channel, not a task | No finish line exists |
| The postponed no | "Let me take a look" | Two conversations instead of one, and a worse second one | You already know the answer |
| The favor delivered as a job | "Don't worry about it, I'll swing by" | Full delivery obligation with none of the process | No paperwork, no slot, no number |
| The seat | "Sure, I'll sit on that" | A recurring tail between the meetings | You cannot name the exit |
| The take-back | "Just this once, I'll handle it" | A reversed delegation that stays reversed | Someone was already doing it |
| The conversation-ender | "Yeah, fine, we'll do it" | An obligation bought to end a discussion | It was late in the day |
Open-ended availability is the most expensive because it has no unit. Every other shape is at least a countable thing. This one converts a single interaction into a permanent channel, and the person on the other end is not abusing it - they are using exactly what you offered. The correction is not to withdraw it, which reads as a rebuke. It is to replace an unbounded offer with a bounded one at the next natural moment: a named window, a named person who answers first, and your line kept for the case that actually needs you.
The postponed no costs about double a clean no, and the second conversation is worse than the first would have been, because now there is a period of expectation behind it. Owners choose it for a real reason: an immediate no in front of someone feels like a rejection of the person. It is not kindness though, it is deferral with interest, and the interest is paid by the other person's planning.
The favor delivered as a job is the one with teeth beyond the hours. Work performed outside your normal process is work performed without the estimate, the scheduling slot, the documentation, the paperwork trail and often the coverage that your ordinary process exists to provide. The hours are the small part. The tell is that you did not ask any of the questions you ask every other time.
The seat on a board, a committee, an association group. The meeting is the visible cost and it is the minority of the real one. The real cost is the tail: the emails between meetings, the thing that needs a call, the person who now has a reason to contact you directly. A seat is also the hardest shape to leave, because the exit requires a story and you did not agree an end date at the start. Take seats deliberately, with a term you state out loud when you accept, and the shape stops being a trap.
The take-back is the only shape that damages someone else's job. You reclaim work already delegated, once, for a defensible reason. What the other person learns is not "he helped." It is "this class of thing routes to him under pressure," and the next one routes to you without anyone deciding it should. This is the shape that turns an owner back into the bottleneck fastest, and it always feels like the responsible choice at the time.
The conversation-ender is a decision-fatigue artefact. It clusters late in the day, at the end of long calls, and at the tail of meetings that ran over. You are not agreeing to the thing, you are buying the end of the discussion. The tell is retrospective relief: you felt lighter the second you said it, which is the opposite of how agreeing to real work feels.
The tell that runs under all six
Four questions, and any one of them failing is enough to slow down. You are not looking for a no, you are looking for a pause long enough to make the yes real.
- Can I name the finish line? Not the task, the finish. "Until they are set up" is not a finish line, "one call and one written summary" is.
- Can I name the hours to the nearest half hour? If your estimate is a shrug, the estimate is zero and you will discover the real number in installments.
- Can I name who does it if I am out for two weeks? If the honest answer is nobody, you just added a single point of failure to a shop you are trying to make less dependent on you.
- Did I answer in under ten seconds? Any commitment worth making survives ten seconds of silence. The ones that do not survive it are the ones being sold on momentum.
The practical version is one sentence you can say to anyone, in any of the six situations, without giving offence: let me check what I already owe this week and come back to you today. It is not a stall if you actually come back the same day. It converts a reflex into a decision, and it is the only intervention on this card that works while the conversation is still happening.
A worked reconstruction: one take-back, six weeks
An owner had moved permit and approval chasing to his office lead four months earlier. On a Tuesday a submission came back rejected on a technicality, the office lead was mid-quote, and he said the words: leave it, I'll handle this one.
His estimate at that moment, had anyone asked, was about 40 minutes. Call it 0.67 hours.
What it actually ran, tracked over the following six weeks:
- The initial correction and resubmission: 20 minutes.
- Four check-back exchanges with the approving office, roughly 15 minutes each: 60 minutes.
- One further resubmission after a second technicality: 50 minutes.
That is 130 minutes, or about 2.2 hours, against an estimate of 0.67 hours. Already 3.2 times the estimate, and that is only the first-order cost.
The second-order cost is the one the tell was pointing at. Over the next six weeks, two more approval items arrived, and both were routed to him without discussion, because the last one had been his. Those cost about 35 minutes each, 70 minutes in total. Add them and the true figure is 200 minutes, about 3.3 hours, against a 0.67-hour estimate. Five times.
Run it back through the four questions as they stood on the Tuesday. Finish line: he could have named it, resubmission accepted. Hours: he would have said 40 minutes and been wrong by a factor of five, but he would have said a number. Who does it if I am out for two weeks: the office lead, who had been doing it for four months, which is the question that should have stopped him. Under ten seconds: yes, comfortably.
Two of four failed, and the two that failed were the two that predicted the actual damage. Note what the questions do not do: they would not have told him the job would run 3.2 times his estimate. Nothing available on that Tuesday would have. What they would have told him is that he was routing work around a person who already owned it, which is the finding that mattered and the one he could have acted on with a single sentence: bring it back to me when you are out of the quote, and if it needs a call I will make the call.
Two commitments that look bad and are not
Over-correcting here is a real failure and it produces an owner who is technically well defended and slowly becoming irrelevant. Two shapes get flagged by the tell and should survive it.
The deliberate investment with a named end. Time spent on something with no near-term return can be entirely correct - training a lead tech, a relationship you are building on purpose, learning a part of the business you have avoided. What separates it from the bad version is that you can state an end condition. "Two ride-alongs a month until he runs the callback list alone" passes. "Spending more time with the crew" is a mood, not a commitment, and it will expand to fill whatever is available.
The one that costs more this quarter to cost less next quarter. Documenting a process, building the answer to a question you get weekly, sitting with someone while they do the thing badly. These run over their estimate almost by definition, and running over is not evidence of a mistake. The test is whether the hours are meant to come back. If the same hours will be spent again next quarter in the same shape, it was not an investment, it was a subsidy you have started paying quietly.
References
- U.S. Small Business Administration (SBA), small business owner workload and delegation guidance
- Trade-standard practice, owner commitment management in small field-service shops
- See related: How to Audit What You Agreed To Last Month, How to Get Out of a Commitment Gracefully, Saying No Without Burning the Bridge, How to Say No to Work That Does Not Fit, The Owner Who Is the Bottleneck and Doesn't Know It