Primary data, on the other hand, are structured by the researcher for the purpose of conducting experimental ad hoc research (research focussed on a narrow and specific objective). Such structuring depends on the researcher's knowledge, intuition and experience for its validity. Validity in this context refers to the capacity of a measurement device, such as a questionnaire item, to measure what the experimenter intends it to measure, without prior bias or verbal misunderstanding.
Since collection of primary data is ever more expensive and respondents ever more difficult to recruit, samples tend to be quite small and sometimes fail to be properly representative of the population in general. While many well designed, adequately representative and validly measured primary surveys are completed, their use in directing marketing actions at a small geographic level is limited or impossible since their small sample sizes render them incapable of generating statistically significant results for specific areas. The smallest areas for which national survey respondents can be tabulated hold about 100,000 persons. These are the Public Use Microdata Areas of the United States Census of Population and Housing (PUMAS). This data source will be discussed in more detail below.