It Worked There, It'll Work Here
Precedent is not evidence. The conditions that produced it are, and they usually stay behind.
Bryan J. Kaus
To copy an example of success, without understanding it with the aid of theory, may lead to disaster.
- W. Edwards Deming
At the General Motors assembly plant in Fremont, California, cars used to come off the end of the line useless, unable to move. They had to be towed away to be fixed. That was considered normal.
By the early 1980s Fremont had the worst vehicle quality in General Motors, a history of wildcat strikes, absenteeism that made staffing a daily improvisation, and a relationship with the United Auto Workers that had stopped resembling a negotiation. GM closed it.
In 1984 the same building reopened as NUMMI, a joint venture with Toyota, with Toyota managing the operation. Roughly 85 percent of the start-up workforce were UAW members from the plant that had just failed. Another five to ten percent came from a shuttered Ford facility down the road in Milpitas. The people GM management had privately regarded as the problem, including the men with reputations as agitators, walked back through the same gate.
Inside of a year the plant went from the worst quality in General Motors to the best.
Same building. Same workers. Same union. The thing that changed was the production and management system.
The lesson everyone drew, and the one that mattered
The obvious conclusion was that GM should copy Toyota, and GM was in an unusually good position to do it. This was not industrial espionage or reverse engineering from the outside. GM owned half of NUMMI. It could send its own managers through the plant, and it did.
It still took the better part of two decades to move much of that capability into the rest of the company, and by most accounts the transfer was partial at best.
The usual explanation is that GM’s culture resisted it, which is true and also unhelpful, because culture is the name we give to a mechanism we have not bothered to better define and specify. The more useful account is that what traveled from Japan to Fremont was not a set of practices. It was a set of conditions that made those practices rational for the people actually executing them.
An Andon cord that lets a line worker stop production only functions if management has genuinely accepted the cost of a stopped line and will not punish the person who pulled it. Standardized work only improves anything if the people doing the work own the standard and can change it. Small-lot production without buffer inventory requires strong supplier relationships and equipment reliability that take years to build and that expose you badly if either fails. Problem-solving at the line requires that surfacing a defect be safe for the person who found it.
Toyota did not ship a toolkit to Fremont. It shipped the playbook and the conditions, and it spent enormous effort installing them, and the tools then worked.
What GM subsequently tried to move into its other plants was the visible layer. The cords, the boards, the terminology, the kaizen events. Those artifacts were photographable and could be installed quickly. The conditions underneath them could not, and without the conditions the artifacts were theatre, and the workforce identified them as theatre immediately, because workforces always do.
The precedent traveled once and failed to travel the second time, and the same company was on both ends of it.
Sixteen samples
I read across a lot of sectors, on the theory that the way things actually work tends to repeat from one to the next, and the clearest version of an idea often turns up a long way from where you need it. This one turned up in a medical journal.
Most of what the world knows about Ebola comes from two strains. There is a third, called Bundibugyo, different enough that the people studying it were candid that the existing knowledge might not carry over. And the entire evidence base for it, across every outbreak on record, is sixteen samples.
Sixteen. That is what a response team works from when an outbreak starts and the decisions cannot wait. The only real body of knowledge available is about two related but genuinely different viruses, and nobody gets to postpone the call until better evidence shows up.
That is not a failure of science. Rather, it is the ordinary condition of any consequential decision, and it is where an operator or an investor lives most days.
What this costs in business
The transfer error is the most common analytical mistake I encounter, and it almost never looks like a mistake while it is being made. It looks like pattern recognition, which is what we pay experienced people for.
Margins from one geography imported into another where the competitive structure and conditions are entirely different. Adoption curves observed among early users applied to a mass market that wants a different thing for different reasons. A management team that performed brilliantly at one scale, in one industry, in one capital environment, hired on the assumption that the performance was the person rather than the fit. Commodity behavior from one regime carried into another where the marginal supplier has changed or economics are different - think conventional oil and gas versus renewables; I’ve written on this exact point before. Returns generated across a decade of nearly free capital extrapolated into an environment where money costs something. Project economics from one power market moved into another with a different market design, a different queue, and a different regulator.
In each case the precedent is real. Somebody did achieve that margin, that adoption curve, that return. The reasoning failure is not fabrication, rather it is the silent assumption that the result was produced by the thing you can see rather than by the surrounding conditions you did not inventory.
The trap is that a precedent is not evidence in the way it feels like evidence. That the past resembles the future, or that this case resembles that one, is not something the numbers prove. It is something we assume, and then arrange the numbers on top of. That does not make experience useless. Experience is information. It makes it a tool with a stated tolerance, and using a tool past its tolerance is how people get hurt and results miss the mark.
So the question to ask of any analogy is not whether a precedent exists. Precedents always exist, and a diligent analyst can find one for nearly any proposition.
The question is which causal conditions produced the precedent, and which of those conditions are present here.
Then, immediately: which are absent, and do the absent ones strengthen the case, weaken it, or break it entirely.
Where this argument breaks
The failure mode of everything above is an organization that has learned to say “our situation is different” and has thereby immunized itself against all outside evidence and blind belief in its own exceptionalism.
That sentence is the single most common defense of a bad practice inside a struggling business. Every plant is different. Every market is different. Every customer relationship is unique. All of which is true and none of which is a reason to keep doing something that works nowhere. Demanding perfect transferability before acting is not rigor, but rather, it is polite paralysis, and the people who practice it usually get to be right about the risk and wrong about the outcome.
The right posture is the one those same doctors took. Faced with sixteen samples, they did not throw up their hands and they did not pretend the gap was not there. They used what they had, said plainly where they expected it to fail, and built a way to find out fast from the people closest to the outbreak.
That has three parts.
Use the imperfect precedent, because it is what you have.
Name precisely where you expect it to fail.
And build the mechanism that will tell you which, fast, from the people on the ground.
The last part is the one many companies skip. Headquarters tends to see the pattern across the portfolio and mistakes that for seeing the mechanism. The plant manager, the field engineer, the regional commercial lead, and the crew running the equipment know which conditions are actually present in a way no comparative analysis will surface. And I say this as someone who started their career in a plant. Excluding them is not merely a discourtesy. It removes the only instrument capable of measuring whether the analogy holds and the principles can be deployed effectively.
The Point Taken
Experience is worth a great deal, and it is worth almost nothing in the form most people carry it. What most of us store is outcomes. What transfers is mechanisms, and the two get filed in the same drawer.
Three questions before you rely on any precedent, whether it is a case study, a prior success of your own, or the thing that worked at the last company.
What were the conditions that produced that result, stated specifically enough that someone could check whether they exist here.
Which of those conditions are absent in this case, and is any single absence sufficient to break it.
And who on the ground would know first if the analogy is failing, and what would have to be true for them to tell you in time.
© 2026 23.5 Strategies; The Point Taken™