There are a number of possible methods of undertaking primary research:
It should be noted that these approaches are not mutually exclusive and more than one may be appropriate for a specific piece of primary research, depending upon the aims of the research, the resources available, the various target groups and their geographic locations.
Focus groups are particularly useful in drawing out issues within a group format, although there are issues about group dynamics which must be considered. Running good focus groups is a lot harder than it appears. The issues identified in focus groups provide useful detailed information although a question must remain as to their overall representativeness of the population at large. One possible method to overcome this is to combine this approach with one of the survey approaches, using the survey approach to find the views of the larger population for comparison to the focus group.
For any primary research, the value of a small pilot piece of work cannot be overstated. A pilot, even if it only involves a handful of possible respondents, can highlight ambiguous questions or missing areas of enquiry which can be rectified prior to undertaking the entire piece of research.
Selecting appropriate questions and designing questionnaires to maximise response is an important technical skill to possess. A number of good tips are contained in the SPSS Survey booklet referred to at the end of this guidance note.
In terms of the survey methods identified, a brief review of their benefits and drawbacks is provided below:
|
Mail survey |
Telephone survey |
Face-to-Face survey |
Cost of data collection |
Low / Medium |
Medium |
High |
Likely response rate |
Moderate / Low |
Moderate / High |
High |
Data collection timescale |
Long |
Short |
Long |
Questionnaire detail |
Moderate |
Low |
High |
Response to sensitive issues |
Moderate |
Moderate |
High |
Question complexity |
Moderate |
Simple |
Complex |
Opening up issues |
Low |
Moderate |
High |
Interviewer bias |
None |
Low |
Moderate / High |
This summarises the benefits of the approaches identified, with the issue of costs remaining to the forefront. It should be reiterated that more than one approach could be adopted when undertaking primary research, and the table above highlights the importance of selecting the appropriate method. Of course although selecting the appropriate method(s) underpins successful research, it does not automatically guarantee success. Each method must still be skilfully and carefully executed to guarantee the quality of results.
One of the first issues to be addressed when it has been decided that primary research is required is the selection of an appropriate sample. The group of interest for the primary research is referred to as the population (although this is used in a wider sense than the demographic meaning). A sample must be chosen from this population, and this is done by means of selection from a sampling frame. As an example, the population of interest may be employers in an area, and the sampling frame used may be a proprietary business directory, such as the Yellow Pages business database. When the sample chosen matches the entire population, this is known as a census.
In choosing the sample for analysis, this can be done with a random element (known as probability sampling) or a degree of judgement can be exercised (purposive sampling). The latter has the drawback that the information gathered and the conclusions drawn are unlikely to represent the values of the entire population. In probability sampling, each member of the sampling frame has an equal chance of being selected. This approach includes simple random sampling (choosing at random) or stratified sampling, where the sampling frame is divided into separate groups (strata), and sampling performed randomly within each group (stratum). This is normally done using random number tables or pseudo-random number generators available in most spreadsheet packages. The stratified approach is useful when we already know something about the population under study (e.g. the industrial structure of an area), as it allows this knowledge to be built into the sampling process. We can oversample smaller stratum and correct for this when analysing the results, to make sure that smaller sub-groups are not excluded from the analysis.
Choosing a random sample does not in itself guarantee that the results from a single sample are truly representative of the entire population. The concept of assigning levels of confidence in survey results, or adopting post-survey methods to ensure representativeness are particularly technical, and specialist advice should be sought. What is important, in general, is to ensure that the results are representative of the population under scrutiny.
Poor sampling techniques can undermine and invalidate primary research. Appropriate sampling design is a necessary, but not sufficient, condition in ensuring good quality primary research.
Last Updated: March 2001