Decisions about sample size should take into consideration the size of the target population you are researching (e.g. all residents in a particular community, members of a particular club or association, people in a particular occupation etc.) and the level of accuracy you require from the research.

There are two terms of key importance: the confidence interval and the confidence level. In the following explanation we will use a fictitious example of a survey of library users to illustrate the points made. The confidence interval is a plus or minus figure that indicates how much your research findings may vary from the results you would have got if you had surveyed everyone in the target population.

In our example, we are surveying a random sample of registered library users. The population to which we will want to generalise our findings is all registered users of the library. In our example, the survey has been conducted using a confidence interval of 5, an interval commonly used in survey research.

The survey results indicate that 60% of library users are very satisfied with opening hours. In this example we can be "sure" that if we had asked all registered users of the library, between 55% (60 minus 5) and 65% (60 add 5) would have said they were very satisfied with opening hours.

The confidence level is expressed as a percentage and refers to how "sure" you can be that true responses would lie within the confidence interval stated. In this case, how sure can we be that between 55% and 65% of library users would say they are very satisfied with opening hours if we had surveyed all registered users?

Our example of research with library users adopted a 95% confidence level, commonly applied in survey research. We can therefore say that we are 95% sure that between 55% and 65% of people using the library are very satisfied with opening hours.

CAUTION: The above calculations assume that you have a truly random sample and cannot be used in circumstances where non-random sampling techniques are used, or where there are flaws in the execution of samples that result in consistent bias e.g. using business club/association members as a sample base for a random survey of "all businesses", or undertaking a general household survey on week days between the hours of 9.00 am to 5.00 pm, thus excluding a large proportion of the working population.