Reversible and Irreversible Decisions in a Small Shop
Why this matters
Most owners spend their decision energy evenly, and that is the mistake. A price test that can be undone next month gets weeks of agonising, while a five-year lease gets signed in ten days because the landlord wanted an answer. The fix is one question asked before any other: if this turns out wrong, what does it cost to walk back, and how long does that take? Decisions that are cheap and fast to reverse deserve speed. Decisions that are expensive or impossible to reverse deserve care. Getting that sort right matters more than getting any single decision right.
The language: one-way and two-way doors
The framing that stuck comes from a 2015 Amazon shareholder letter. A two-way door is a decision you can walk back through if you do not like the room: try it, watch, reverse. A one-way door closes behind you: once through, the old option is gone or only recoverable at heavy cost.
The useful refinement for a small shop is that reversibility is not yes or no. It has two measures:
- Reversal cost: what you lose by undoing it, in money, hours, customers or trust.
- Reversal time: how long it takes before you can undo it and be back where you started.
A decision with a reversal cost of a few hours and a reversal time of a week is a two-way door. A decision with a reversal cost of four years of rent is a one-way door even though, technically, you could break the lease.
Where common shop decisions sit
| Decision | Reversal cost | Reversal time | Door |
|---|---|---|---|
| Raise one service's price 5% | A few lost quotes, or a discount to win them back | Next price sheet, days | Two-way |
| New dispatch rule (e.g., first-call slot held for emergencies) | Some office confusion | A week | Two-way |
| Try a new supplier for one product line | Part returns, a restocking fee | One order cycle | Two-way |
| Hire a tech | Training hours already spent, a crew that watched you hire badly | 1 to 3 months | Mostly two-way |
| Buy a truck outright | Resale loss, typically a meaningful share of price in year one | Weeks to sell | Partly one-way |
| Sign a 5-year lease | Remaining rent, or a negotiated exit | Years | One-way |
| Fire a long-tenured tech | Their customer relationships, crew trust | Rarely recoverable | One-way |
| Drop a commercial account that is a fifth of your work | The account, and the reference it gave you | Usually never | One-way |
| Public promise ("no price increase this year", "lifetime labor warranty") | Your word, broken in public | Never cleanly | One-way |
Notice the last row. Promises made to customers or staff are among the most irreversible decisions a shop makes, and they are often made in a sentence, on the phone, with no thought at all.
The two errors, worked through one shop
A plumbing and HVAC shop with 9 people made both errors in the same spring.
Error one: treating a two-way door as one-way. The owner wanted to raise maintenance plan prices 6%. He worried about cancellations, asked three peers, built a spreadsheet, and debated it for 7 weeks. The shop renews about 25 plans a week. During those 7 weeks, 7 times 25 = 175 plans renewed at the old price, each locked for twelve months. The forgone increase is 175 times 6%, which is the revenue of about 10.5 plans. When the increase finally went in, cancellations ran no higher than the shop's usual rate. The cost of the delay was real and the decision could have been reversed in a day by offering anyone who complained the old rate.
Error two: treating a one-way door as two-way. In the same spring the owner signed a 5-year lease on a building with twice the bay space, in 10 days, because the landlord had another interested tenant. The reasoning was "we can always sublet." Growth stalled in month 8 when a lead tech left. Subletting half a shop building with shared bays turned out to be nearly impossible, and the shop carried 52 more months of rent on space it used half of.
The same spring, run the other way round. The price change becomes a two-week test: the 6% goes on renewals starting Monday, the office logs every cancellation and every complaint, and the owner sets a reversal line in advance (say, cancellations running more than double the usual weekly rate). At 25 renewals a week, two weeks is 50 renewals, enough to see a spike if there is one and few enough that refunding the difference to all 50 would cost the increase on just 3 plans (50 times 6%). The lease gets the deliberation instead: the owner asks for a 2-year term with a renewal option, accepts a slightly higher rate for it, and when the lead tech leaves in month 8 he is carrying 16 months of exposure rather than 52.
The pattern is common and it has a cause. Two-way doors feel risky because the downside is vivid and immediate (an angry customer on the phone tomorrow). One-way doors often feel safe because the downside is distant and abstract (rent in year four). Your attention goes to the wrong one.
Things that look reversible and are not
- Anything that changes what people believe about you. You can rehire a tech you fired in anger, but the rest of the crew has already updated what they think job security means here. You can raise a price you cut, but the customers who got the cut now anchor on it.
- Deleting records or history. Purging old customer records, switching software without exporting job history, letting a phone number lapse. Recovering these is often impossible.
- Exits from scarce relationships. Leaving a supplier who gives you allocation in a shortage, or a landlord in a tight market. The door you walk out of may not be there when you come back.
- Decisions taken in a cash crunch. A reversal that needs money becomes one-way when you have none. When cash is short, classify everything one notch more irreversible than usual. See related: Scarcity and the Decisions a Cash Crunch Makes.
Things that look irreversible and are not
- Price increases. They feel permanent and are among the easiest decisions to adjust, because you control the next quote.
- Software and process changes. Switching field software is months of pain, but you can run old and new side by side, and the pain is mostly a one-time cost, not a closed door.
- A new service offered as a test. If you sell it to a small group and call it a trial, stopping is cheap. If you buy dedicated equipment and advertise it citywide first, it is not.
- Most hires, in the narrow legal sense. Employment is at-will by default in every state except Montana, whose Wrongful Discharge from Employment Act requires good cause once a probationary period ends, and other states carve exceptions through implied contracts and public policy. So the legal reversibility of a hire depends on your state and your paperwork; take a specific termination to an employment attorney with the offer letter and handbook in hand. This is not legal advice. The social cost covered above applies everywhere.
Converting a one-way door into a two-way one
The most valuable move is often not deciding faster or slower but changing the decision's shape so it can be undone. Each of these has a price, and the price is usually worth paying:
- Lease terms: a 2-year term with a renewal option, or a break clause at month 24 with a few months' penalty, instead of 5 years flat. Ask before you sign; landlords often agree for a slightly higher rate.
- Rent or lease equipment before buying. A jetter, a lift, a second truck for one busy season.
- Temp-to-hire or a defined trial period for a new role, with the review date written down.
- Pilot before rollout. One tech, one zip code, one product line, one month.
- Announce changes as tests. "We're trying Saturday hours through August" can be withdrawn without breaking a promise. "We're now open Saturdays" cannot.
For the hardest one-way doors, run a premortem and check the base rate for decisions like it before you sign. See related: Running a Premortem Before a Big Commitment, Base Rates: The Number You Should Start From.
Speed rules for two-way doors
The cost of treating every decision as one-way is not just the delay on that decision. It is the queue: an owner who agonises over a price test also sits on the dispatch rule, the supplier trial and the uniform order, and the office learns that nothing moves without him.
A working default:
- Reversible within a week at low cost: decide within a day, or let whoever is closest decide.
- Reversible within a month: decide within a week, with a written review date.
- Reversible only over months or at real cost: this is where your deliberation belongs.
Two-way doors are also the ones to hand off. If a decision can be cheaply undone, the person nearest the work should usually make it. See related: Decision Rights: Who in the Shop Decides What.
What changes the classification
- Scale. Raising one service's price is two-way; repricing the whole book the week before the busy season is closer to one-way, because the season's work is booked at whatever you choose.
- Timing. A hire in March gives you the shoulder season to find out; a hire in July gives you no slack to reverse before the peak.
- Who else is committed. A decision is less reversible the more other people act on it. Once the crew has reorganised schedules around a four-day week, rolling it back is a one-way door for them even if it is two-way on paper.
When you are unsure which door you are standing at, ask what the decision would cost to undo in 90 days, and write the answer in hours, months or customers. If you can write a small number, move. If you cannot write a number at all, slow down until you can.
References
- Amazon, 2015 Letter to Shareholders, on one-way and two-way door decisions
- Montana Wrongful Discharge from Employment Act, Mont. Code Ann. 39-2-901 et seq.
- See related: Keeping a Decision Journal, Running a Premortem Before a Big Commitment
- See related: Decision Rights: Who in the Shop Decides What, Scarcity and the Decisions a Cash Crunch Makes