In a survey, a group of people are invited to answer a series of questions. Most often, survey questions ask respondents to select an answer among several predetermined options (e.g. very satisfied, somewhat satisfied, or not at all satisfied), but questions can also be open-ended to allow for greater feedback. Surveys can range in length from several minutes to an hour or more, and the number of people who take part in surveys varies greatly.

Most surveys consist of many questions. However, if only a few questions need to be asked, a good option is an omnibus survey (in which a limited number of questions are purchased on a longer survey).

The results from well-conducted surveys can be projected onto a population larger than the people who took part in the study. For example, if a survey found that 60% of women exercise at least once a week, it can reasonably be projected that 60% of all similar women (plus or minus several percentage points) would exercise. Survey error (the plus or minus) is related to several factors, three of which are described below.

The most well-known component of survey error relates to the size of the sample. Simply stated, the more people who take part in a survey, the less error there is in projecting survey results to a larger population. However, there are diminishing returns as sample sizes get larger in reducing this error. (Many national surveys set a size of around 1,000 as a reasonable trade-off between error due to sample size and cost.)

A less well-known, but potentially more meaningful, source of survey error is related to what is called the sampling frame. The sampling frame is the universe of people who could be invited to participate in the survey (e.g. all visitors to a Web site), from which a sample is selected. Survey results can only be only as accurate as the sampling frame, so it is important that the frame accurately represents the population of interest.

An additional source of survey error is related to response rates. Not everyone who is invited to take part in a survey will participate. Higher responses rates usually mean less survey error.

Surveys can be conducted via telephone, mail, over the Internet, in person, or some mixture of these methods. How the survey is conducted impacts cost, the time it takes to conduct the study, and survey error.

Uncommon Insights conducts all types of surveys, including bilingual surveys in Spanish and in English. Please contact us for more information on our survey capabilities.