The marketing sample

The sample, as we have seen, is a small group taken from the total group (the universe). This total can be a city, a state, a nation, the whole world, or it can be a relatively small universe, such as the total number of students enrolled in a high school for the current academic year.
A good marketing sample has the following four main qualifications which must be met if the researcher is to use his findings to project them to the total universe.
(a) It must be random. By "random," we mean that everybody in the group has an equal chance of being chosen as a respondent such as, for example, every tenth student in the high school as they come out after school. (But note that it would not be random if we took every tenth student who comes out for football practice, since many students do not come out for football, and naturally all girls would be excluded. )
(b)    It must be representative. By "representative" we mean that the sample must include all important kinds of units in the total universe. For example, we must include representatives from each of the three junior high and three senior high grades. We must include both boys and girls, of all ages, and of all economic groups.
(c)    It must be proportional. "Being proportional" means the sample contains the various segments enumerated under representativeness in approximately the same proportions as exist in the total universe. In a typical city high school, for example, we might find sixty per cent white, forty per cent non-white. Our sample would have to reflect those proportions. Further, we might find 85 per cent native and 15 per cent foreign. These proportions would also have to be reflected in our sample.
It must be adequate. Since we are selecting a given number out of a total universe, we must be careful to choose an adequate number so that we can say with assurance that the results of the findings from the small group represent the total universe. Naturally, the larger the sample, the more nearly correct such a full representation will be. But in all sampling work, it is found that, after a rather limited number, there is a tendency to level out. In many studies, such a leveling out or repetition tends to show itself after even 100 or 200 replies.
Suppose, for example, we were to ask 1,783 students how many of them would like to see rugby substituted for football in their school. We could ask the entire 1,783 students (the entire universe) or we could make up a representative random sample of say 250 students. After the first 100, there would tend to be a repetitive pattern, and later additions would simply confirm this. One such study in an eastern high school showed the following:
First 100 83 no, 15 yes, 2 don't know Next 100 81 no, 14 yes, 5 don't know Third 100 82 no, 15 yes, 3 don't know
It was safe to conclude that four out of every five students in the school did not favor the substitution of rugby for American-style football.
The danger in all sampling is bias. Bias often results from lack of information, lack of understanding, lack of proportionality of responses (even when the sample was set up properly), mistake in interpretation due to limited experience or limited technical knowledge on the part of the researcher. Bias can result from any shortcomings of the sample, of the interviewer, of the respondents, and most especially, of the interpreter. Distortion can easily lead to greatly distorted conclusions.