The IQR rule flags outliers using quartiles, so it does not depend on the mean or the standard deviation at all.
That makes it usable on data of any shape, not only on bell-shaped data.
It needs only sorting and counting, so it is easy to compute and easy to explain.
Step 1 is to find the first and third quartiles, the values 25% and 75% of the way through the sorted data.
Step 2 is the IQR itself: the third quartile minus the first.
Step 3 sets the lower fence 1.5 IQRs below the first quartile and the upper fence 1.5 IQRs above the third.
Any value outside the fences is flagged for a closer look.
For a normal distribution the fences sit about 2.7 standard deviations from the mean, so very few values are flagged.
Using 3 IQRs instead of 1.5 gives outer fences, which flag only the most extreme values.
The rule ignores the shape inside the fences, so 2 clusters with a gap between them pass without a flag.
For strongly skewed data, many genuine values in the long tail still fall outside the upper fence.
Taking logarithms before applying the rule often gives a more sensible set of flags for incomes or prices.
A lower fence below what is possible, such as negative minutes, simply means no low values can be flagged.
In practice the rule is a good first screen for every numeric column during profiling.
The usual mistake is treating 1.5 as a law, when it is a convention that can be tuned to the data.