Sampling Bias

Sampling bias occurs when the way you select your sample makes some members of the population more likely to be included than others, so the sample no longer mirrors the whole. The result is a skewed picture that no amount of careful math can later repair.

A famous case: a 1936 magazine poll predicted the wrong U.S. president because it sampled phone and car owners, who were wealthier than typical voters. A modern echo is judging customer satisfaction only from the people who answered a survey — the contented and the furious answer more often than everyone else. The defence is random or otherwise representative sampling, where every member of the population has a known, fair chance of being chosen.