The Jobs Worth Walking Away From

Why this matters

Most walk-away advice is about margin, difficult people and gut feel, and all three are poor predictors. Shops turn down good work because a customer was short on the phone, and take on the job that eats a quarter because it priced well. The jobs that actually hurt share one property and it has nothing to do with how they felt: they contain an unknown that cannot be handed to anybody. Not to you, not to the customer, not to a supplier, not to an insurer. An unknown with no owner is carried by whoever is standing closest when it lands, and on your own job that is always you.

The measure

Every unknown in a job has to end up in one of four states before you commit a price:

• Resolved. You performed some act and now you know.
• Excluded. Stated in writing as outside scope, and the customer still gets what they paid for without it.
• Covered by an allowance. A stated quantity is in the price, with a rule for truing up past it.
• Unassigned. Nobody owns it.

The test for walking away is not how many unknowns a job has. It is whether any of them are stuck in the fourth state after you have genuinely tried the first three. A job with nine unknowns, all assignable, is a normal job. A job with one that will not move is the one to leave.

The list

This is a list and not a ranking. There is no honest measure that sorts these against each other: the same entry that ends a small job is survivable on a large one, and their frequency varies more by trade than by anything intrinsic. Anyone presenting them as ordered by cost is decorating.

1. A scope defined by an outcome you cannot measure. "Make it quiet." "Make it comfortable." "Make it not smell." You can measure sound pressure, temperature and airflow; you cannot measure quiet. The unknown here is the acceptance criterion itself, which means the customer holds a test you cannot pass on purpose. Assignable when: you convert the outcome into a measured target both of you agree to in writing before the price.

2. An unknown behind something nobody will open. Concealed piping, buried conduit, a chase, a slab, a sealed cavity. The unknown is unobservable, and if the deliverable depends on what is in there it is also unexcludable. Assignable when: someone with authority permits an exploratory opening before you price, or the deliverable is redefined so the concealed section stops mattering.

3. A dependency on a party you have no agreement with. A landlord, another trade, a utility, an equipment supplier with no confirmed date, an inspector whose interpretation is the real gate. You are being asked to promise a completion that a stranger controls. Assignable when: you get their commitment in writing, or your agreement makes their performance a condition rather than your risk.

4. A customer who will not let you document what you find. No photographs, no written findings on the invoice, no conditions listed. This one is disqualifying in proportion to how much you will discover: harmless on a like-for-like swap, fatal on anything exploratory, because the entire mechanism for assigning a discovered unknown has been removed in advance.

5. A payment structure where you finance the unknown. Retention released on a completion somebody else defines, progress payments tied to a milestone that depends on entry 3, or a final payment gated on an approval nobody has scheduled. The unknown is not technical here, it is whether the trigger ever occurs. Assignable when: the release trigger is an event you control or a date.

6. A safety condition the customer intends to keep. A defeated interlock they want back, a hazard they will not have corrected, a system they intend to keep running in a state you have identified as dangerous. This one is different from the other five: it is not that the unknown is unassignable, it is that the outcome is known and unacceptable. Walk, and the written finding still goes to them on the way out. See related: How to Refuse a Job Without Losing the Customer.

What is not on this list, and why

This is the part worth reading twice, because these are the reasons shops actually give.

Thin margin. A job that prices at a margin you find unattractive is a pricing decision, not a risk decision. If you can deliver it for the hours you quoted, the worst case is that you earned less than you wanted. That is a different category of bad from a job that cannot be finished. Plenty of comfortable-looking jobs belong on the list above and plenty of thin ones do not.

Old equipment. Age correlates with unknowns and is not one. A 30-year-old system that is fully accessible has fewer unassignable unknowns than a new one buried in a chase. Judging by age is how shops decline the accessible old job and accept the concealed new one.

A customer who was difficult on the phone. Difficulty is a customer question and it deserves the customer-refusal treatment, which is a different instrument with a different shape. Some of the most demanding customers document better than anyone.

Work outside your comfort but inside your competence. A gap you can close by training, by subcontracting to somebody you have an agreement with, or by buying an hour of somebody's expertise is a resource question. It only becomes entry 3 if the person you would rely on has made you no commitment.

A long payment cycle. Slow is a cash question and it can be planned for. Entry 5 is not about slow, it is about a trigger that may never fire. Those two get confused constantly and they need different responses.

Two jobs that look identical

A mechanical shop bids two commercial kitchen ventilation projects in the same month. On the estimate sheet they are near twins: occupied restaurants, similar equipment, tight schedules, both quoted at about 120 hours.

Both carry the same unknown: a section of exhaust duct has to run through a concealed vertical chase, and nobody knows what is in it. If the chase is clear, that section is roughly 8 hours. If it is obstructed and the run has to be re-routed, it is up to about 40 hours. That is a swing of 32 hours on a 120-hour job, about 27% of the total.

Job A. The building owner is the customer and agrees to a 2.0-hour exploratory opening before pricing. The chase is opened, it is clear, the 8 hour figure holds and the unknown is gone. Two hours of observation removed a 32-hour uncertainty, which is 16 hours of swing eliminated per hour spent looking. Treat that as an information ratio, not a return: the 2.0 hours are real cost and the 32 hours were never going to be spent anyway. What it bought was a number the shop could stand behind.

Job B. The customer is a tenant. The landlord will not permit any opening before the work starts and the tenant has no ability to compel one. Run the four states. Not resolvable, because nobody will grant access. Not excludable, because the deliverable is a ventilation system that has to pass inspection and the duct has to get where it is going. Allowance? This is where most shops go wrong, so look at the failure shape rather than the size.

A 27% swing is priceable. An allowance with a true-up handles it, if the customer agrees. What kills Job B is that the failure mode is not a swing at all. If the chase is obstructed, the only alternative route crosses a demising wall the same landlord controls, and permission for that is no more available than permission to look. The realistic bad outcome is not 40 hours instead of 8. It is a job that cannot be completed at any hour count, with a restaurant partly torn up and a shop with no contractual path to either finish or stop.

That is the distinction the list turns on. A range you can price is not a reason to walk. An outcome that may not exist is. Entry 3 is the actual entry Job B trips, and entry 2 is the symptom.

Job B is also not permanently disqualified. It becomes assignable the day the landlord signs anything, or the day the agreement makes landlord access a condition precedent to the shop's schedule and price rather than a hope. A shop that walks and says exactly that out loud is often the shop the tenant calls back with a signed letter three weeks later.

What flips an entry off the list

Every entry above except the sixth is a condition rather than a verdict, and the flip is usually cheap:

• Entry 1 flips on a measured acceptance criterion agreed before price.
• Entry 2 flips on access granted by whoever has authority to grant it, which is not always the person paying you.
• Entry 3 flips on a written commitment, or on a contract term making that party's performance a condition rather than your exposure. Anti-indemnity and pass-through provisions vary substantially by state and some states restrict what a contract can shift onto a subcontractor, so have your own attorney read the terms you are being asked to accept rather than deciding from a card.
• Entry 4 flips on a scope with nothing discoverable in it.
• Entry 5 flips on a release trigger that is a date or an event you control.

Entry 6 does not flip. A customer's intention to keep a dangerous condition is not an unknown you can bound.

How to verify you are reading a job right

• Write the unknowns down before you write the price. If you cannot list them, you are not evaluating risk, you are reacting to a smell.
• Mark each one with one of the four states and refuse to leave any blank. A blank is the finding.
• When you decline, name which entry you declined on. A shop that declines six jobs a quarter and cannot say which entry each one tripped is declining on feel and calling it discipline.
• Track the ones you took with an unassigned unknown. Every shop takes some. What you want to know a year later is whether the unassigned ones are over-represented in your loss-making jobs, and that comparison only exists if you recorded the state at bid time rather than reconstructing it afterward.
• Check your declines against the negative-space list. If most of your walk-aways cite margin, age or a difficult customer, you are declining the wrong jobs and accepting the ones on the real list.

References

• See related: How to Decide Whether to Work on a System You Did Not Install
• See related: The Walk-Away Power: Knowing Your Number
• See related: How to Refuse a Job Without Losing the Customer
• See related: Reading a Contract Someone Else Wrote Before You Sign It