When designing research projects, it’s essential to use a sample that is representative and large enough to be robust. For general opinion polling or consumer research, this usually means potential sample sizes, are only confined by budgetary and/or timing limitations, as the total number of people to draw a sample from (which statisticians call, “the universe”) is very large.
However, exceptions to this occur when the probability (also know as “the incidence rate”) of finding the type of person that the research is specifically interested in, by random sampling alone, is low. Examples of target samples with a low incidence rate would include, consumers of a very niche product or service, a very specific type of person based on their demographics, behaviours or attitudes, or managers at certain levels of seniority who work in specific types of companies. These groups of people would all have a relatively finite universe size, and therefore it is unrealistic to set a target sample size that is too high, not only based on cost and time limits, but also what the ‘real life’ universe size dictates is feasible for market research.
This is particularly appropriate for business-to-business research, which effectively aims to capture perceptions of organisations, by surveying relevant decision-makers in different companies. For example, the manufacturer of a new employee payroll system, may want to find out what its target market thinks about it; in this instance it would be appropriate to survey anyone involved in the procurement of such a system, in companies large enough to consider purchasing it. Based on the number of Private Sector Medium/Large organisations (defined as 50+ employees) in the UK, which latest ONS figures estimate to be just over 40,000, it would be unrealistic to expect to be able to survey more than a few hundred of their HR/Board level Directors, as it is simply not possible to reach enough of them, to generate a larger sample size. However, that wouldn’t make results any less unreliable; the sample is robust enough to represent the views of the right types of companies, allowing conclusions about all of these organisations to be made, with confidence.
So when setting a target sample size, consideration should be given to not only cost/time limitations, but also what its universe size suggests is feasible.