Surveyors sometimes cannot arrange a probabilistic sample and instead rely on a variety of non-probabilistic techniques, each which poses potential problems.
Surveyors could: target a quota of a certain type of respondent and then move on to other types; construct a sample of convenience based on volunteers or captive audiences; a priori restrict the sample to individuals with predefined characteristics; or use snowball sampling, also called network, chain, or reputational sampling, where follow-on contacts are provided by those sampled earlier. Each of these non-probabilistic methods opens the door to biased and inconsistent results.