Sunk Cost in a Bad Hire, a Bad Truck and a Bad Customer

Why this matters

The three most expensive things a small shop keeps too long are a tech who is not going to make it, a vehicle that keeps eating money, and a customer who costs more than they pay. In each case the owner can usually say exactly how much has already gone in: the weeks of ride-alongs, the transmission, the six years of discounts. That number feels like a reason to stay. It is not, and it can also push the other way, into dumping something that is actually fine because it once burned you. This card gives you the questions that separate what is already spent from what is still in play, and runs them on all three.

The idea, in business form

A sunk cost is money, time or effort already spent that you cannot get back whatever you decide next. The sunk cost effect is the measurable tendency to keep investing in something because of what has already gone in. Hal Arkes and Catherine Blumer showed it in 1985 with people who kept going to a ski trip they no longer wanted because they had paid more for it. The diagnostic version, where a tech keeps chasing a theory because of the hour already spent on it, is covered elsewhere; this card is about the slower, bigger version that plays out over months.

The rule is short: only the future goes on the ledger. The cost already paid appears on both sides of every option, keep or exit, so it cancels. What decides is what each option costs and returns from today forward.

The trap is not that owners do not know this. It is that a sunk cost rarely shows up labeled. It shows up as "we have too much in him to quit now", "we just rebuilt that van", or "they have been with us since the beginning."

Why a shop is especially exposed

You made the call yourself. In a large company the person who bought the truck is not the person who sells it. In a shop of eight, both are you. Walking away reads, to you, as admitting the first decision was wrong, and that is a cost to your pride that no ledger records.

The investment has a face. The hire is someone you trained. The customer is someone whose kids you know. The van is number 3 and everybody calls it that.

Real future costs hide inside the sunk ones. This is what makes the trap hard: not everything in the "already spent" pile is irrelevant. A replacement hire takes weeks to recruit and months to ramp. Selling a van means a gap in the schedule. Losing an account means filling those hours. Those are future costs and they belong on the ledger. The discipline is sorting each item into the right pile, not throwing out everything that sounds like history.

The five questions that cut through it

Run these in order. Each one strips out a different piece of the history.

1. The fresh-start question. If you did not already employ this person, own this vehicle or serve this account, would you take it on today, on exactly these terms, knowing what you know now? If the honest answer is no, the only remaining argument for keeping it is switching cost, and you can measure that.
2. The forward ledger. Over a stated horizon, what does each option cost and return from today? Write both columns in hours, days down, callbacks or percentage of margin. Nothing that has already happened goes in either column.
3. The successor test. If someone bought the shop tomorrow and walked in fresh, what would they do in week one? A new owner has no pride invested and no memory of the transmission bill. Andy Grove described using this exact question at Intel to drop a business its founders were attached to.
4. The salvage question. What part of what you spent is still recoverable? Resale value on a vehicle is real. Training a tech received is real to him but not to you once he leaves. A customer's referral network may survive the account or may not.
5. The deadline question. If you decide to keep going, what would you need to see, by what date, to keep going after that? A keep decision without this is just postponing the same conversation. Setting it is covered in its own card; here it only needs a date and a number.

Case one: the hire who is not coming along

A shop of nine hired a tech fourteen weeks ago. So far the lead tech has spent about 120 hours on ride-alongs and checking his work, and the shop paid for a certification class. His callback rate is the problem: 3 of his 20 jobs the month before last came back, and 3 of 22 last month. The shop's own rate runs about 1 in 25.

The owner's instinct: "We are 120 hours into him. Give it another few months."

The fresh-start question: would you hire him today, knowing his callback rate has not moved in six weeks? The owner says no.

The forward ledger, in hours, with each assumption stated:

Over twelve months, keeping him costs 88 more hours than replacing him (298 minus 210). The 120 hours already spent appear in neither column.

What changes the answer: the horizon. Run the same table over six months and it flips. Keeping him costs 45 hours of callbacks (18 at 2.5) plus 104 hours of checking (26 weeks at 4) for 149 hours. Replacing costs 40 plus 85 plus 60 plus about 12.5 hours of callbacks for roughly 197.5 hours. Over six months, keeping is about 48 hours cheaper. Replacement is front-loaded and pays back later, which is exactly why owners who think in "a few more months" increments keep the wrong person for two years.

What else changes it: the trend. If his callbacks had run 5 of 20, then 4 of 21, then 3 of 22, the forward column would be shrinking month by month, and keeping him with a deadline would be the right call. Flat at a bad number is different from improving toward a good one.

The legal fork. Most US states default to at-will employment, meaning either side can end it without cause; Montana is the exception, where its Wrongful Discharge from Employment Act requires good cause once a probationary period ends. At-will does not protect a termination that is retaliatory or discriminatory, and that is a federal and state question. If this tech recently complained about pay or safety, took protected leave or filed a workers' comp claim, stop and talk to an employment attorney before acting, with his file, the callback records and the dates. This is not legal advice.

Case two: the van you just rebuilt

Van 3 is nine years old. Four months ago the shop put a transmission in it, at roughly a sixth of what a comparable used replacement would cost. Now a mechanic says the rear main seal is weeping and the suspension needs work, together about a third of the transmission bill.

Here the sunk cost can push either way, and this time the gate says keep.

The owner's first instinct is the reverse sunk-cost reaction: "I already threw a transmission at that thing, I am not putting another cent in it." That feels like discipline. It is the same error, because it lets a past bill decide.

The fresh-start question: if you could buy this exact van today, with a four-month-old transmission and a known seal and suspension job, would you? Run the forward ledger over two years. The repair is about a third of the transmission bill, which puts it well under a tenth of a replacement. The van was down 6 days in the last 12 months, 5 of them for the transmission, now done. A replacement means a week or more of shopping, upfitting shelving and moving stock, and the replacement carries its own unknown history.

So the forward cost of keeping is one moderate repair on a van whose biggest known failure is behind it. Keep it, and set the deadline question: if van 3 is down more than 5 more days in the next 12 months for anything other than routine service, it goes.

What changes the answer: a frame-rust finding, a second major drivetrain failure, or downtime that has started costing booked jobs rather than just shuffled ones. Each of those moves the forward column, not the backward one.

Case three: the account that has been with you since the start

A property manager has used the shop for six years and accounts for about 8 percent of revenue. The relationship started with a discount for volume. Since then they pay at around 75 days against your 30-day terms, three of last quarter's twelve invoices were disputed over scope the tech had confirmed on site, and the office spends roughly 5 hours a month chasing paperwork for them. The margin on their work, after the discount, runs about two thirds of the shop's average.

"They have been with us since we had one truck" is the sunk cost talking. Six years of loyalty is real and is also entirely in the past.

The fresh-start question: a new property manager calls tomorrow offering this exact volume at this exact discount, paying at 75 days and disputing a quarter of the invoices. Would you sign? The owner says no.

The successor test: a new owner would reprice the account or end it in the first month.

The forward choice here is rarely "fire them today." It is usually reprice and re-paper first: standard rates, the agreement's payment terms enforced, scope signed on site. If they accept, the account becomes one you would sign fresh. If they leave, you have 8 percent of revenue to replace, and the question becomes whether you can refill those hours at a better margin, which most shops with a waiting list can within a season.

If there is a written service agreement, its notice and termination terms govern how you exit, and reading them before you say anything to the customer is the first move. Where the agreement is unclear, take it to your own attorney before giving notice.

What the money already spent is still good for

Sunk does not mean useless. Three things carry forward legitimately:

• Information. The 120 hours of ride-alongs told you something about how this shop trains and whom it hires. That knowledge is worth using in the next hire, not in deciding about this one.
• Salvage. A rebuilt transmission raises the van's resale value. Count the resale, not the bill.
• Reputation. How you end a relationship is visible to the rest of the crew and to the customer's network. That is a future effect and it belongs on the ledger.

Two ways this goes wrong that the questions do not catch

The ritual "one more chance" with no number attached. An owner who keeps a struggling hire "until things settle down" has not decided anything. Without the deadline question answered in writing, a keep decision silently renews itself every month. See related: Setting Kill Criteria Before You Start Something New.

Overcorrecting into ruthlessness. An owner who has just learned about sunk costs can start dumping anything with a past investment to prove he is rational, which is the van case above wearing a different hat. The questions are symmetric: they can say keep, and in practice they say keep about as often as they say go.

References

• Arkes, H. R., and Blumer, C. (1985), "The psychology of sunk cost," Organizational Behavior and Human Decision Processes
• Grove, A., Only the Paranoid Survive (1996), on the new-owner question
• Montana Wrongful Discharge from Employment Act, Mont. Code Ann. 39-2-901 and following
• See related: The Sunk Cost Trap in a Long Diagnostic Session; Setting Kill Criteria Before You Start Something New; Reversible and Irreversible Decisions in a Small Shop; Loss Aversion and Why Owners Will Not Raise Prices