2014: Advent Quality: Risky Widgets

by on December 12, 2014

The art of risk assessment is arcane, mysterious, and vital to any sane testing regimen.

You are the widget-maker. You are a responsible and safety-conscious widget-maker, who wishes their widgets to be trustworthy. You intend to implement a testing procedure, to ensure that your widgets meet the standards you claim. How many should you test?

There is no right answer here, and by that I don’t mean “any answer you like is correct”. I mean that you are responsible if your widgets fail, regardless of how well thought through your testing plan was. It is your responsibility to sell only safe widgets.

So, pragmatically, you do a risk assessment. You look at your processes in detail and you decide where errors and flaws are likely to happen, and pay more attention to controlling those parts of the process. You might test the first five widgets in a run, to ensure that the machines are configured correctly. You might test every fiftieth widget, or three widgets every hour. You might test every single widget for a simple non-destructive thing like size, by building a filter into your production line to remove the oversized ones.

You also look at how bad the results will be if the widget fails in a particular way before you determine how often you’ll test for it. It’s much worse for your widgets to hurt people than for them to fail safely; it’s worse for them to fail safely than for them to be the wrong colour.

Leave a Reply