Know all about representativeness in surveys

‍

When conducting a study, choosing the population to interview is a crucial step, as it will determine the predictability of the results you will collect.

If your study focuses on a shampoo brand, for example, you will want to interview shampoo buyers. However, you do not have the time or the resources to interview all shampoo buyers. You will therefore select a sample, that is to say a small number of individuals who are part of your target. Generally, we choose to select a sample whose profile is the same as that of the target population, this is what we call a representative sample.

Why select a representative sample?

‍If your sample is representative of the total population, you will be able to extrapolate the results of your sample to the total population.

Concretely, if your sample is representative and if 20% of the shampoo buyers in your sample say they want to buy your new product, then you can assume that 20% of the shampoo buyers who see your new product in store when it is launched will want to buy it.

This generalization will allow you to operationalize the results for your future strategy.

How do I know if my sample is representative?

‍A sample is said to be representative if its profile is identical to that of the population studied. The profile variables used to build the sample must be chosen according to the population studied and the analyses that one wishes to carry out. In market research, sociodemographic variables such as gender, age and socio-professional category are generally used, but it is entirely possible to add a variable specific to your target audience.

The important thing is to clearly define the variables that can have an impact on the results. For example, if you think that buying shampoos in supermarkets or rather in pharmacies can have an impact on the attractiveness of your new product, it will be necessary to have a representative sample on this variable.Framing your sample on the shampoo purchasing circuit will prevent you from having an imbalance between the share of supermarket buyers and the share of pharmacy buyers in your study vs. in real life, which could distort the results of your study.

‍Representativeness is not related to your sample size

‍As we have just seen, the representativeness of your sample is linked to its profile. Representativeness is therefore not linked to the size of the sample interviewed. You don't have to interview a large number of people for your sample to be representative. A sample of 100 people can be just as representative as a sample of 1000 people.

‍There are several methods for obtaining a representative sample

‍The most common are those of the random sample or the quota method. It is the latter method that is generally used in market studies, or marketing studies. It involves setting the percentages of individuals to be interviewed for each profile variable chosen. Indeed, if you want to interview a sample of 1000 individuals representative of the French population and if, according to INSEE, the French population is composed of 49% men and 51% of women; then, you will have to interview 490 men and 510 women.

Profile questions are asked at the beginning of the questionnaire in order to quickly determine whether the respondent corresponds to the required profile and can respond to the study or not.

In order to streamline and ensure the quality of the collection of interviews, a tolerance is generally applied to quotas. Indeed, if a single interview is missing from the study but according to the quotas this respondent must be between 18 and 35 years old and retired, it will be very complicated to find this type of profile in the population and if it exists it would be too atypical to wish to keep it in the study.

Setting up a quota tolerance makes it possible not to block the completion of the field. For example here, it would allow an individual aged 18-35 to answer the questionnaire even if he is not retired. To correct the discrepancies associated with this tolerance, the data is adjusted when the field is closed. The aim is to apply a weighting of the data in order to model the profile of the sample on the profile of the total population to ensure its representativeness. Representativeness is therefore not a matter of sample size but rather of proportions concerning the profiles interviewed. To be sure to interview a quality sample, representative of your target, do not hesitate to contact us!