Editor’s note: Bonnie Eisenfeld is a self-employed marketing research consultant based in Philadelphia.

A common question in the beginning of a survey research project is, how large should the sample be? There are precise mathematical methods to answer this question but often budget considerations are equally important. A statistician may specify the sample size but the research project budget cannot accommodate that many. Are there good arguments for a budget increase? Will the research be useless if the sample is too small?

Using the correct size sample increases the probability that your findings accurately reflect the target population. If you understand the purpose of correct sampling, you can make better decisions about the sample size to ensure that research outcomes will be dependable.

  • The sample should be large enough so that you see important patterns. Suppose you are looking at fabric to recover a sofa, and you get a one-square-inch swatch. You may not see the pattern of large flowers in the fabric. If you get a three-foot-by-three-foot swatch, you will be able to view the pattern. Similarly, in market research, you need a sample that is large enough so that you are confident that it reflects the pattern of behavior or attitudes in the target population.
  • The sample should be large enough so that you can see significant differences between segments. If you are looking at two pieces of fabric, each one-square-inch in size, that are only slightly different in color, you may not see the difference, but you surely will see the difference between the two colors if you compare three-foot-by-three-foot swatches. Similarly, in a research sample, if you are comparing segments of the population, you need a sample size that is large so that you are confident you are seeing any significant differences between segments of the population.
  • The method for obtaining the sample is just as important as the sample size. Suppose your fabric has a border that is different from the middle. You need to get a swatch that shows both sections and best represents the entire piece of material. If you looked at only the border swatch, you would miss the main pattern. Similarly, in a research sample, you need to avoid bias in sampling to ensure that each member of the target population has an equal chance of being drawn for the sample. By avoiding bias in the sample, you will be confident that your conclusions are based on the behavior or opinions of the entire target population and not skewed towards only part of it.

A common problem with sampling bias in qualitative research is called a convenience sample. This situation occurs frequently when recruiters produce repeaters or their friends and not fresh participants. Recruiters can also produce a biased sample if they use a list that has a built-in bias that excludes certain people. And some less ethical recruiters will produce bias by accepting ineligible participants.

  • The sample does not have to be proportional to the population. When obtaining fabric swatches, you would not need to get larger swatches from the larger rolls of fabric and smaller swatches from the smaller rolls. Similarly, in research samples, it is often possible to draw sound conclusions using the same-size samples from a population of 1,000 as from a population of one million.
  • Respondents are more difficult to obtain than fabric. The analogy between research samples and fabric swatches ultimately doesn’t work because people do not always want to be in the sample, whereas fabric does not object. The main reason respondents participate in surveys is their interest in the topic. To ensure a high response rate, you can explain the purpose of the survey, the importance of their participation, and how the findings will be used. Incentives for participation will increase response rate; the higher the incentive, the better the response, in most cases.

The list from which the sample will be obtained needs to be large enough to account for non-participation. If you could be sure that 100 percent of those people you asked to participate would do so, you wouldn’t need to start with such a large list.

  • There is always some chance of sampling error, but in a correct-size sample, the chance of error is minimized. Sampling error means that the sample is slightly off the mark. Maybe you got a fabric swatch from the manufacturer on a bad day and it had flaws. Maybe it was cut from a flawed part of the roll. In a research project, using a larger size sample means a smaller sampling error, which is what you want to ensure precision in your findings.
  • If you need advice on sample size for a research project, ask a statistician not a decorator. But don’t hesitate to demand an explanation of terminology in plain English.

Â