Researchers begin with an idea of what they want to study, for example the moisture content of cheese. The researcher then selects a "population," in this case which exact cheeses to study. The sample must accurately represent the characteristics of the larger population of cheeses or the results cannot be generalized. For example, the researcher can't simply pick one type of cheese because one type of cheese is not representative of all cheeses.
Primary, secondary and tertiary units are how a population is broken down into its constituent elements. For example, in a study of the effects of an insecticide spray, the primary unit of study is a particular plot of trees. The secondary units are the trees within the plot. The tertiary units are the apples on the trees.
In a household survey of a particular city, the primary sampling unit is a particular block in the city. The secondary sampling unit is the particular households chosen for the survey. The tertiary sampling unit is the individual chosen to answer the survey question. For example, a researcher may call the chosen household within a particular block of a city and ask to speak to the mother of the house for the purposes of conducting a survey. The mother is the tertiary sampling unit.
A researcher who wants to conduct a pilot project has to choose the population upon which to test the project. For example, a researcher wants to test the effectiveness of a cognitive skills program in rehabilitating young offenders. The researcher needs to select the primary sampling unit, here the locations for testing the program. The secondary sampling unit is the specific institutions and the tertiary sampling unit is the individual offenders chosen for the program.