Market surveys are carried out for various purposes such as estimating the need/pain point for a particular product, gauging public opinion for improving a product etc. Preparing an unbiased sample is one of the biggest challenges before conducting a survey Many marketers and researchers alike have known to falter while choosing the sample for their survey. But once sample is determined, the next question which comes to mind is how many people need to be surveyed? One might say get as many participants as you can. But is a “large” sample really needed? After all a large sample will require many resources to be employed which will increase the cost and also delay the size of the results of the survey.
Sample size depends upon four parameters:
- Population size
- Confidence interval (allowable margin error between sample mean and population mean)
- Confidence level (How much confident one wants to be that the sample mean falls within the confidence interval)
- Standard deviation σ
Confidence interval and confidence level are generally expressed in percentage. many sample size calculators are available online which can be used to determine the sample size. It is generally observed that sample size increases when confidence interval and confidence level are increased i.e in order to reduce errors, a bigger sample is needed. The most interesting observation is that when population size is increased sample size increases. But after one point, for a given confidence interval and confidence level, an increase in population size does not lead to an increase in sample size which leads to a conclusion that there is an upper limit on the sample size. Though this may sound a bit odd there is a logical explanation behind this observation. As sample size increases initially the responses received tend to vary a lot. But after a point, the responses become homogeneous which is logical as a sample is made up of similar elements. So an increase in sample size after a point will not lead to a drastic change in the sample statistics. The results from an online sample size calculator (shown below) also point to this fact.
A research based survey design can be broadly classified into three categories:
Exploratory research is carried out when there is a lack of clear idea of the problem. It is characterized by high uncertainty. Focus group discussions, Delphi techniques, projective technique, case studies and experience surveys are a few methods of carrying out exploratory research. Focus group discussion involves a discussion among twelve to fifteen individuals on a problem. Four to five people observe the body language and note down key discussion points. Delphi technique involves forming an panel of experts and discussing the research problem with them. Delphi technique is an iterative process. Projective technique involves the use of thematics, word associations, sentence completions etc. to gauge the opinions of people. The results of projective techniques should be interpreted by a trained psychiatrist.
Descriptive research is carried out when level of uncertainty is medium. Descriptive research can be broadly classified into two types, longitudinal surveys (surveys done repeatedly over a period of time) and cross sectional surveys (done once a while). Census, credit ratings of financial instruments, readership surveys carried out by IRS and NRS are a few examples of longitudinal surveys.
Causal surveys may be deterministic or probabilistic.