Training in the FMEA method

I'm sitting here in a training class in Sticksville on the FMEA methodology, and the instructor just finished introducing a typology of failures. Briefly, any kind of tool, product, or service is defined by having a certain kind of function. And once you know what the function is, then a "failure" is a deviation from the desired function. And there are five possible kinds of deviation:

• No function at all ("total failure")
• Quantitative deviation (it does the right thing but in the wrong amount, too little or too much)
• Time deviation (it does the right thing but too early, or too late, or it misfires somehow)
• There is an unintended or undesirable function
• There are impermissible side effects (noise, heat, radiation)

So far, so good. But it got me thinking: do these same types apply to human failures as well? I think they do ... and here is my first pass at thinking through how that could work.

“How many kinds of failure are there, now?”

The trainer asked us, reading from the screen.
And answered, “Five — it all depends on how
“The non-fulfillment of the function’s seen.”

“There’s total failure: nothing’s done at all.
“Then quantitative: little or too much;
“Early or late; wrong functions great or small;
“Or side-effects, like noise and heat and such.”

I harbor all of these deep in my heart:
The sluggard who just dreams, I know him well;
The one who does it backwards, end to start,
Who’s thoughtless, and whose voice clangs like a bell.

For every type of failure we define
Is surely one that I can claim as mine.