In my last post on quantitative and qualitative research, I describe the two different types of research conducted by firms (not forgetting primary and secondary research). With these types of research, the firms have to pick people to conduct the research on. This is known as sampling which is what this article will be on: Random sample, Quota sample, Stratified sample and Sample sizes.
Sampling is a way of ensuring that results are typical of the whole population, though only a sample of the population can be used.
Now with sampling, there are a few concerns in how to choose the ‘right’ people (sampling method) to carry out the research on and decide how large a number to use (sample size). The ‘right’ people being the type of people that can portray the views and opinions of the whole population. There are three main sampling methods:
This involves selecting respondents to ensure that everyone in the population has an equal chance of being picked. It sounds easy but its not. If an interviewer goes to a street corner one morning and asks passers-by for an interview, the resulting sample will be biased towards those who are not at work, who do not own a car and have time on their hands (the busy ones will refuse to be interviewed). As a result the sample will not be representative of the whole population. So achieving a truly random sample requires careful thought.
Research companies use the following methods:
- Pick names at random from the electoral register (e.g. every 50th name)
- Send an interviewer to the address given in the register
- If the person is out, visit up to twice more before giving up (this is to maximise the chances of catching those who lead busy social lives and are therefore rarely at home).
The method above is effective, but slow and expensive.
This is selecting interviewees in proportion to the consumer profile within your target market. A good example of this appears below, where the percentage of buyers for each gender of chocolate reflects the respondent quote:
|An example of Quota sampling
This method of sampling allows interviewers to head for busy street corners interviewing whoever comes along. As long as they end up achieving the correct quota, they can interview when and where they want to. As this is a relatively cheap and effective way of sampling, it is the one used most commonly by market research companies.
This method of sampling involves interviewing only those with a key characteristic required for the sample. For example, the producers of Olay might decide only to interview women aged 30-45, the potential buyers of the future. Within this stratum/section of the population, individuals could be found at random (hence stratified random sample) or by setting quotas based on factors such as social class and region.
After deciding what type of sampling method a firm uses, they will then have to decide their sample size: 10, 100, 1000? What sample size should you use?
Of course, if you interviewed only ten people, the chances are slim that the views of this sample will match those of the whole population. Of these ten, seven may say they would definitely buy Orange Chocolate Buttons. If you asked another ten, however, only three may say the same. A sample of ten is so small that chance variations makes the results meaningless. In other words, a researcher can have no statistical confidence in the findings from a sample of ten.
A sample of 100 is far more meaningful. It is not enough to feel the confidence about marginal decisions (53% like this 47% don’t), but is quite enough if the result is clear-cut (75% like this to the 25% that don’t). Many major product launches have proceeded research on as low a sample as 100.
With a sample of 1000 a high level of confidence is possible. Even small differences would be statistically significant with such a large sample. So why doesn’t everyone use samples of 1000s? The answer is money. Hiring a market research agency to undertake a survey on 100 people would cost approximately £10,000. A sample of 1000 people would cost three times that amount – good value if you can afford it, but not everyone can. As shown in the earlier example of launching Orange Chocolate Buttons, a company might require six surveys before launching a new product. So the spending on research alone might reach £180,000 if samples of 1000 were used.
Backdata – keeping records of the results from past research to provide a comparison with the latest results.
Bias – a factor that causes research findings to be unrepresentative of the whole population.
Primary research – finding out information first-hand
Secondary research – finding out information that has already been gathered.
Sample Size – the number of people interviewed; this should be large enough to give confidence that the findings are representative of the whole population.
Sampling method – the approach chosen to select the right people to be part of the research sample.
Random Sample is random
Quota sampling when selecting people in proportional with the market
Stratified is using people to interview that have key characteristics