Unbiased sample
An unbiased sample is one whose makeup does not differ systematically from the population it represents, so its estimates land on the true value on average, with no consistent lean in one direction. The term describes how the sample was selected; any single result can still miss by chance.
How it works
Bias enters samples in several ways. Selection bias arises when the method favors some people, such as surveying only website visitors. Coverage error arises when the frame misses part of the population. Nonresponse bias arises when the people who take part differ from those who decline. Self-selection bias arises when participants volunteer.
Random selection from a complete frame is the main defense, because it takes the researcher's choices out of who is included. Researchers then compare the achieved sample with known population figures and use weighting to correct remaining imbalances. Sample size does not help on its own, since a larger sample drawn the same way carries the same bias.
A worked example
Four thousand customers answer a retailer's pop-up satisfaction survey after online checkout, and 88 percent say they are satisfied. A random sample of 800 drawn from the full customer database, in-store shoppers included, and contacted by email and phone, finds 71 percent. The pop-up heard only online shoppers, and only those who chose to answer straight after a purchase went through.
Related
Automate this in Pulse
The screener in a Pulse brief is where a team can check that no group is excluded by accident, one of the ways bias enters a sample before the first interview.