Multiple Library Customer Service Survey My office has purchased a customer satisfaction survey for 106 public libraries.  There are 189355 active library members in the province but only 677 individual chose to complete the survey.  Would this be considered an accurate survey to make policy for libraries across the province.  (44 library systems had zero responses. 43 branches had less than 5 responses for the year.)
 A: It depends on how the survey was conducted. If 700 people were sampled at random from a database of library members and 677 completed the survey, then the survey may well provide a good picture of the population. With random sampling, you will just by chance get a sample that doesn't include all the libraries; that's ok.
What raises a problem is if the people in that sample are not there through random selection (or something equivalent). If the survey was conducted by just haphazardly leaving some English-language questionnaires at the library checkout where some 677 bored, civic-minded people filled them out, then the survey is probably not providing an accurate picture of the population. For example, you might miss out on library members with low English proficiency, or you might miss out on library branches where questionnaire distribution was especially sloppy. If civic-minded people are the only ones taking the survey, then you'll get overestimates of things like how many members volunteer at the library.
In short, whether your survey provides an accurate picture of your population depends on:
(1) which people were given an opportunity to participate in the survey; and
(2) which of those people actually did complete the survey.
To have any guarantee of your survey's accuracy, you almost always have to control (1) through random sampling. In addition, to avoid issues caused by (2), you have to have either very high response rates or a convincing nonresponse bias analysis that indicates survey respondents are similar to nonrespondents in the things you're trying to measure in your survey.
