Mistake and Cause

The popular cult of the mistake suggests celebrating every misstep as a free lesson. Mistakes are unavoidable, but they don’t become useful automatically.

After a mistake, a person knows that one method didn’t work under particular conditions. They don’t necessarily know the right method yet. The number of ways to ruin a result is close to infinite, and working through each of them systematically would take too long.

Still, a mistake can yield important information if you take it apart:

• what was expected;
• what happened;
• where the divergence appeared;
• what the immediate cause was;
• what conditions allowed it to take effect;
• why the mistake wasn’t caught earlier;
• what will change the system, not just the current result.

Here it matters to distinguish a mistake, an experiment and negligence.

A mistake happens during honest work under incomplete knowledge. An experiment admits an uncertain result in advance and limits the consequences. Negligence means known precautions were ignored.

The right to make mistakes doesn’t include the right to step on the same person’s foot repeatedly and call it an educational process.

Hiding a mistake is usually more dangerous than admitting it. While the bad news belongs to you, you can report it together with an assessment of the consequences and a plan of action. Once others discover it, the discussion will no longer be only about the mistake.

Failures shouldn’t be romanticised either. The line “mistakes are the best teacher” sounds suspiciously like an advertisement for a teacher with very high fees.

It is worth studying not only failures and successes, but also near misses. They often reveal a weakness in the system before any damage appears.

For a mistake to teach anything, recording the method that failed isn’t enough. You have to go from the event to the cause and to the conditions for its recurrence.

Why did the machine stop? The safety cut-out tripped. Why did it trip? The motor overheated. Why did it overheat? A bearing was creating extra resistance. Why wasn’t the bearing replaced? Checking it wasn’t in the maintenance schedule. Why wasn’t it? The schedule was copied from a different model.

If you stop at the first answer, the solution is to reset the cut-out. The machine will stop again, but this time with a sense of a job well done.

A sequence of “why” questions helps you move from the event to the conditions that created it. You don’t have to ask exactly five times. Sometimes two are enough. Sometimes, after the eighth, it turns out the investigation has quietly moved on to the origin of the universe.

There can also be several causes. An accident rarely happens purely because of one person, one screw or one bad environment. Usually a technical defect, an organisational weakness, unclear responsibility and a well-chosen moment coincide, when all the protective barriers happened to be looking the other way at once.

A deep explanation shouldn’t merely sound convincing. It should indicate what will change once the cause is removed.

This search matters most where the usual fix removes the discomfort but leaves the mechanism of the problem intact.

A person is constantly late and sets the alarm earlier. Then earlier still. A month later he has six alarms, chronic sleep deprivation and the same lateness. Perhaps the problem isn’t the time of the signal, but going to bed late, an overloaded morning, an unrealistic estimate of the journey, or a job he doesn’t want to turn up to.

It is worth separating:

• the symptom;
• the immediate cause;
• the conditions that let the cause recur;
• the system of incentives around the problem.

Problems are especially durable when their subject matter is simultaneously somebody’s profession, budget, status or source of power. This doesn’t mean every specialist has an interest in keeping the problem alive. But a system can reward handling the consequences more than removing the causes.

The best solution isn’t necessarily counterintuitive, and it is rarely magically simple. More often it just changes the conditions under which the problem kept reproducing.

Sometimes a repair is enough. Sometimes a new process is needed. Sometimes what’s needed is to stop treating the symptom as a problem. Sometimes it turns out there is no full solution and the damage has to be reduced instead.

In a kitchen this plays out plainly. A knife left blade-up in a sink full of water hasn’t cut anyone yet. Formally nothing happened, there is nothing to report, and by the end of the shift nobody will remember it. The cut will happen a week later, to a different person, on a different shift — and it will be recorded as their inattention. The injured person is the one who gets reviewed, because the event belongs only to them, while the habit is spread across the whole shift and belongs to nobody in particular.

Reviewing near misses is more expensive: there are many of them, there is no damage, and there is no appetite for the discussion. But at that stage the conversation is still about the sink, not about who is to blame.