Sears built its reputation on DieHard batteries and Craftsman tools, the kind of dependable, no-nonsense brand a customer trusted without thinking twice. In the early 1990s, that trust started leaking out through the auto centre.
The company had changed how it paid its mechanics, tying pay to quotas with commission stacked on top. Sell more, earn more.
The logic held up in a boardroom. A repair shop paid for volume would produce volume.
California's Department of Consumer Affairs ran an undercover operation between 1990 and 1992. Investigators brought in cars that needed nothing more than a simple brake job. In 34 of 37 visits, Sears staff recommended repairs the cars didn't need, some running up bills of $550 over what the job should have cost.
The fallout spread fast. New Jersey, Florida, Illinois, and New York opened their own inquiries, auto centre revenue fell 15% within months, and Sears scrapped the commission structure and settled with attorneys general in 41 states.
Nobody at Sears set out to defraud a customer. They built a system that rewarded the technician for finding more to fix, and every technician did exactly what the system paid them to do.
Most decisions get one test before they're approved: will this work. That test has an answer built in, because the decision was designed to make the answer yes.
It only measures the decision from the inside.
Howard Marks, the investor who built Oaktree Capital, calls the deeper test second-level thinking. First-level thinking stops at "this is a good idea." Second-level thinking asks what happens once everyone downstream reacts to it.
Sears asked whether commission would lift sales. It would, and it did.
Nobody asked what a mechanic does once the invoice becomes his paycheck.
Before you approve a decision, run it through five questions, not one. What are the downstream consequences once this leaves the room. What happens if it works better than you expect, and what happens if it works worse. What unintended consequences are plausible, even the ones that sound unlikely today.
One question in that list usually does the real work: who benefits, who loses, and how will they respond.
Run it on Sears' pay structure and that question alone would have stopped the policy before it started. A mechanic paid by the repair benefits from finding one. The customer loses if the repair wasn't needed, and a customer who feels oversold doesn't complain quietly.
They stop trusting the shop, then they tell four friends.
The system worked exactly as designed. That was the problem.
Take one incentive or metric driving a decision on your desk this week. Run it through that one question: who loses if this works exactly as intended, and how will they respond.
Write the answer down before you approve anything else.
