Industrial processes such as the production of pills are often controlled by periodically sampling and analysing the product. For instance, one can take every 100th tablet and analyse how much aspirin is actually present. The aspirin content should not be too low, for then the pill does not work, but not too high either because that is too expensive. The results of these analyses are depicted in a so-called "control chart".

This chart is set up by analysing data from a period in which the production process was functioning OK. The variation in aspirin content (a standard deviation) during this period is called sigma. The solid line represents the target aspirin concentration and the dashed lines are the warning limits, plotted at 2 sigma. The dotted lines are called action limits and are plotted at 3 sigma.

What would you do if you were the plant manager and the latest analysis showed up outside the action limits?
Stop all action! Something is clearly wrong.
Call for your boss and let him/her decide. This might be serious.
Do nothing, just wait till the next sample. One in a hundred samples is bound to fall outside the 99% confidence intervals.
Pretend it is not your task to look at little boring details as control charts, and ignore the whole thing.