Sampling bias
Your sample systematically over- or under-represents parts of the population.
What it means
Sampling bias occurs when the method of drawing a sample makes some members of the target population more likely to be included than others, so the sample is not representative. Common forms include convenience sampling, self-selection, undercoverage of hard-to-reach groups, and non-response. The result is estimates that are systematically off for the population of interest, no matter how precisely they are measured. It is distinct from random sampling error: increasing sample size sharpens a biased estimate around the wrong value rather than correcting it.
Examples
A phone poll on landlines once skewed older and wealthier, missing younger, mobile-only voters entirely.
Restaurant ratings skew to the delighted and the furious, because the people who had a perfectly fine meal rarely bother to write it up.
A company's engagement survey looks glowing until someone notices only half the staff replied — and the ones who had already checked out were the least likely to click.
First described in Foundational survey statistics; dramatized by the 1936 Literary Digest poll failure.