It’s easy to think of human error as random and unpredictable but the truth is that lots of study has gone into rates of human error and human reliability and it is far from random — it is intimately connected to the features of our tasks, tools, and environments.
There are some stunning finds if you take a look at the data. My favorites:
- When a system or procedure doesn’t match the reality of the task, the likelihood of error is increased by a factor of 8
- Performing an activity with a shortage of time (rushing) increases the likelihood of error by a factor of 11
- Reading a 5-letter word incorrectly when the resolution of the document is poor results in 3 errors per 100 attempts — compare that with reading a 5-letter word when the resolution of the document is clear: 3 errors in 10,000 — a decrease of a factor of 100
The solutions are even embedded in the lesson — that’s how practical human performance can be sometimes. Make sure work-as-prescribed matches work-as-intended/done. Make sure operators aren’t rushing, particularly during already risky human performance activities. And make sure your documents are clear, concise, and readable. Simple.
Another great learning is related to inspections. You have a 10% chance to fail to recognize an incorrect status during an inspection. You are purposefully looking for problems, and yet there’s a 10% chance you’ll miss something. Compare that to error rates when you have a defined list of things to look for — the rate drops to 0.3%. Have detailed inspection checklists, don’t just rely on folks to catch any and everything.
Now that we know human error isn’t unpredictable, we can begin to understand what error-likely situations are and systematically eliminate them from our processes. Ready, set, go!