Median >>> Average, usually

From analytics curmudgeon Stephen Few's The Data Loom: Weaving Understanding by Thinking Critically and Scientifically with Data, page 80:
Means are appropriate measures of central tendency only when the distribution is normal. When distributions aren't normal, the median is a better measure of central tendency.

Succinct and actionable. As he points out in the book, means (averages) can be particularly skewed by outliers when sample sizes are small. Hopefully this increases your awareness of the potential downsides of using average over median in designing solutions.

The Data Loom is a solid, fairly quick read for anybody who works with data. I believe Few's connecting the dots between the scientific method and analytics will cause me to ask better questions, especially in problem definition. But be forewarned...

Contact Form


Email *

Message *