I get paid to fail 60% of the time. But failing is not a failure.
Every research proposal I wrote contained a small act of honesty hidden inside a sentence nobody noticed.
The sentence usually read something like this: “The team will evaluate five candidate approaches and down-select to two leading contenders.”
To a program manager, that line reads as a plan. To the engineer writing it, the line is a confession. It says, in the politest language available, that three of the five ideas are expected to fail, and that those failures are part of the work rather than a detour from it.
The Tightrope
Much of my 32 years in aviation R&D involved high-temperature composites, materials that live inside jet engines at temperatures that would soften most metals. One recurring question was repair. When a part made from these materials gets damaged, can it be fixed, or does it go in the scrap bin at enormous cost?
A repair program would start with five or so ideas in mind. A bonded patch. A filler worked into the damage. Re-infiltrating the damaged zone. A mechanical insert. And the baseline an honest program carries: replace the part and compare the cost. Somewhere inside, there was usually a favorite. Engineers are human, and humans develop hunches and root for them.
The proposal could not say “idea number three is going to work.” Funding organizations want confidence, and research cannot promise it. Proposals walk a tightrope: confident enough to earn the money, honest enough to admit that most roads lead to dead ends.
“Down-select to two” was the balancing pole.
Run the arithmetic on that sentence. Three out of five is 60 percent. A typical proposal quietly budgeted for a 60 percent failure rate.
What makes that job work is a distinction most people never stop to draw. Failing is an event: the test ran, and the answer was no. A failure is what happens when nobody learns anything from the no.
What a Dead End Is Worth
A good researcher is rewarded not because the first guess is always right, but because experience says which “no” to collect first.
A negative result, the experiment that did not produce the hoped-for outcome, is still a measurement. It records something true about how the material and the physics behave. A patch that lets go at temperature marks a boundary. A filler that cracks defines what the part will and will not tolerate. Each one narrows the search.
Even the most humbling result, “this part cannot be economically repaired,” is worth a great deal. That single sentence can stop a company from spending the next five years on the same question.
Add a Method to the Shelf
Some failures are not dead ends at all. They are side doors.
One repair effort targeted damaged coating on a small, local area of a part. The usual way to prepare a spot like that for a masked recoat is abrasive: blast the zone clean and rough it up so the new coating will grab. Abrasive methods are hard to confine to a small area, so the idea was to replace the grit with chemistry. Leach the bond coat out of only the zone that needed repair, and leave behind a clean, prepared pocket ready for the mask and the new coating.
The leach worked. That was the problem.
It knew no bounds. I can’t go into the chemistry, but the behavior is the point. Once started, it kept going, advancing past the edge of the repair zone into coating it was never invited to touch. No chemical antidote could stop it at the line. As a method for local repair, it failed, and no amount of wishing would move it into the “win” column.
Then the result turned over in my hands. A process that refuses to stop is a liability when the goal is a small, surgical repair. That same process is exactly what you want when the goal is to remove everything. The failed spot-preparation method was, in fact, a total-removal method, ready for the day someone needs a part stripped bare.
Lose, then win. The local repair lost. The shop gained a new way to strip a part completely.
The difference between those two outcomes came down to how the failure was labeled. Filed as “local leach: failed,” the result disappears into a drawer. Filed as “bond coat leach: will not stop at a boundary; candidate for full strip,” it becomes a method on the shelf, waiting for the right problem to walk by. Same data, different label. Write down what a method does, not only what it failed to do for you.
Every Day Is an Education
A phrase carried through my whole career: in R&D, every day is an education, but not every day earns an A+.
School trains us to treat grades as the point, and a report card full of A’s signals a student doing well. In a laboratory, a notebook full of nothing but A’s should make you nervous. It means one of two things: the problems were too easy, or someone is fooling themselves. Real exploration produces C’s and D’s, and the C’s and D’s are where the density of learning is highest.
The B-minus quiz tells you what to study. The failed test specimen tells you where the physics pushed back.
Where Failures Go to Disappear
Successes get written up, which means reports, patents, conference papers, and promotions. Failures get a shrug and a shelf, and too often it is the wrong kind of shelf. Scientists have a name for this: the “file drawer problem,” a term psychologist Robert Rosenthal coined in 1979 for studies with unremarkable results that never leave the researcher’s desk.
In industry, the file drawer runs even deeper. The engineer who knows why approach number four failed carries that knowledge in memory, not in a database. When that engineer retires, the knowledge retires too. Five years later a bright new team proposes approach number four, spends the budget, and rediscovers the same dead end.
That is the most expensive kind of failure: the one already paid for once.
Negative results are the most perishable form of tribal knowledge, because nobody schedules a meeting titled “Things That Didn’t Work.”
Why This Matters Outside the Lab
No lab coat is required to run experiments. The natural instinct is to hide the misses and advertise the hits. The engineer’s instinct runs the other direction. Write down what you tried, what happened, and what you learned, because a failure you understand is worth more than a success you can’t explain. The approach that didn’t work may have taught you exactly what the right one looks like.
Down-Select With Honor
The three ideas that didn’t survive the down-select were never losers. Some were the scaffolding that held the winners up. Without them, the two survivors were just guesses. With them, the survivors were chosen. Others went on the shelf, labeled for what they could do, waiting for a different problem to need them.
Sixty percent of the work failed. None of it was a failure.
So here is my ask. What is the most valuable “failure” in your work or your life, the dead end that turned out to be a side door? Tell me in the comments. I suspect some of the best methods we have are still sitting in file drawers, mislabeled.
Herbert Roberts, P.E. is a licensed professional engineer with 32 years in aviation research and development across three companies, and has spent eight years analyzing accidents under his P.E. license.