Glossary
Survey sampling bias
When respondents systematically differ from the population studied.
Sampling bias occurs when the people who answer a survey are systematically different from the population you actually want to describe, so the results do not generalise no matter how many responses you collect. More responses do not fix a biased sample — they just make the wrong answer more precise.
Common forms in form-collected data: self-selection (only strongly satisfied or strongly angry people bother), non-response bias (the silent majority differs from the vocal minority), and coverage bias (your distribution channel reaches only part of the population, e.g. emailing only existing customers about acquisition).
Mitigations are mostly about design, not size: distribute through channels that reach the whole population, keep the survey short to lift response rate among the indifferent, avoid leading questions, and report who you did not hear from. A representative small sample beats a large skewed one.