Sample Size

Sampling, by design, is imperfect. Sample-based findings are never 100% accurate. They are consequential, however, provided the study is well designed.

The determination of sample size is a commercial decision that weighs the costs of a larger sample against the benefits of greater accuracy. There is not much value in the information sourced from a sample unless it can be generalised to the target population. The ability to do so with some confidence depends on factors associated with sample design as well as non-sampling inaccuracies.

Small unreliable samples that do not permit generalization are not meaningful or useful. Large, overly accurate samples may be needlessly expensive. An ideal sample is one that precisely meets specifications — it is neither over specified nor underspecified. The specification of ideal sample size is dependent on the following factors:

  • Population variability. The larger the variability the larger the sample required to achieve the desired level of accuracy. (Note: For continuous variables such as an item’s averages sales, this variability is best reflected by the relative standard deviation).
  • Sample design. For services like retail audits, a stratified sample design can yield a substantial reduction in sample size.
  • Specified level of accuracy. The standards for sampling error are set by the service provider. The greater the required precision, the larger the sample size.

Other factors specific to the nature of the research also affect sample size. For instance, for retail audits, larger sample of retail stores are required if products are thinly distributed. Similarly, for usage and attitude studies, if product usage is low, a larger sample of consumers will be required to achieve the desired level of accuracy.

It is pertinent to note that sample size is not dependent on universe size. This may sound counterintuitive — if universe size is not a factor, why then do we need large retail audit samples in countries like China and India.

The reason is because variability in these universes is much greater, and product distribution is low. Besides that for large markets, we also have many more market breakdowns, such as regions, provinces, and cities. The sample size for each market breakdown must individually meets the specified accuracy standards, which adds to the total requirement.

To design a sample, at the onset, we need to set the level of precision of the data. This in turn is subject to the nature of the decisions that the research is commissioned to address.

For a wide range of market research programmes, industry (or agency) norms serve as guidelines. Take for instance, Nielsen’s sampling standards in retail measurement.

The level of precision is typically set by specifying the standard deviation (standard error) of the parameter to be estimated, or by stipulating the probability that a particular estimate will be statistically significant.

In addition to the precision levels, the sample size calculation requires knowledge (or assumptions) of the nature of the market. Some of this information is sourced from large scale studies, referred to as household surveys or establishment surveys; for instance, the retail census for retail tracking studies. Other information, for example proportion of population of users of a brand, may be based on conservative assumptions. The calculations, therefore, are inherently hypothetical.

Previous     Next

Note: To find content on MarketingMind type the acronym ‘MM’ followed by your query into the search bar. For example, if you enter ‘mm consumer analytics’ into Chrome’s search bar, relevant pages from MarketingMind will appear in Google’s result pages.