Sampling — common mistakes

In an exam, always link the choice of sampling to business objectives. For example, justify using a small, focused sample for a niche product or a large, stratified sample for a mass-market product launch.

When discussing stratified sampling, specify the 'strata' relevant to the case study, such as age groups, income levels, or geographical regions, to demonstrate application of knowledge.

Justify the use of non-probability sampling in the context of business constraints. A start-up with a limited budget might reasonably use convenience sampling for initial exploratory research, and you should acknowledge this trade-off.

When evaluating market research data in a case study, always question the potential for sampling bias. Ask: Who was surveyed? When? How were they chosen? Could the method have excluded important segments of the population?

Always name the bias (age, location, income, self-selection) when criticising case research.

Not necessarily. While a larger sample size can increase statistical confidence, representativeness is more important. A very large but biased sample (e.g., 10,000 people all from one city) is far less useful than a smaller, well-chosen sample of 500 that accurately reflects the entire population's demographic and geographic mix. The method of selection is often more critical than the sample size alone.

Businesses often face constraints of time and money. Non-probability methods like convenience or quota sampling are significantly cheaper, faster, and easier to implement than probability methods. For initial exploratory research, testing a concept, or when a high degree of statistical accuracy is not the primary goal, the speed and low cost can be a justifiable trade-off against the higher risk of bias.

Both methods involve dividing the population into subgroups. The key difference is how participants are selected from these subgroups. In stratified sampling (a probability method), participants are chosen randomly from each subgroup. In quota sampling (a non-probability method), the researcher selects participants non-randomly until a pre-set quota for each subgroup is filled, often based on convenience. This makes stratified sampling more statistically robust but also more complex and expensive.