What Are Some Of The Most Common Non-Sampling Errors?

Despite the fact that nowadays it is possible to apply research methodologies that reach a large percentage of people, quantitative data collection techniques that use samples are still used in Latin America. As we have mentioned in other posts on this blog, it is an impossible task for most brands to carry out a census to know their consumers. That is why samples are used that are as representative as possible of their population, since at no time can we achieve 100% confidence if we analyze only a part of a universe. The only way to get as close as possible to a knowledge of an entire population is through a census and the coordinated and high-quality work of a team of surveyors.

Ensuring that the sample of people who answer our questions validly is one of the greatest challenges that market research has faced since its inception. The design of the research protocol is as or more decisive than the analysis itself when carrying out a market study. Although there are mathematical mechanisms to know the sampling error of an exercise, these presuppose that the design and data collection phases adhered to the highest standards in the industry.

However, there are no statistical formulas that hold true when inaccuracies are allowed when gathering survey participants. These types of errors are known as «non-sampling» and tend to occur when there is no rigor in the processes, design and team of researchers. They are often dismissed for multiple reasons ranging from hasty execution to unforeseen circumstances. Here are some of the most common non-sample errors.

  1. Geographic representativeness. Many studies seek to know the characteristics of consumers in an entire country or region. However, when sampling by location, they tend to focus on one or a few more populated cities. Although it is known that large Latin American cities dictate the behavior trends of the rest of the inhabitants of the territories under their jurisdiction, it should not be forgotten that in certain cases this will not always guarantee real and generally practicable results. For example, Peru and Mexico are two countries in the region with segments of respondents living in environments very different from those of their capitals. The answers from a person in Mexico City may differ greatly from those provided by someone living in Villahermosa. Likewise, a participant from Lima may have different needs and habits than a native of Pucallpa.
  2. Form is background. In many cases, we believe that the means to perform a data survey is somewhat superfluous. However, choosing one way or another can modify our perception of a phenomenon. For example, in a region where there is low penetration of Internet access and an online survey is launched that seeks to characterize the general population, it will leave out a significant number of people; In any case, it will help us to get to know Internet users better. In these cases, it is preferable to think of a mechanism that allows a more random and universal scope of the respondents. A possible solution may be by placing survey crews in high-traffic sites that invite all passersby to collaborate with our investigation.
  3. Situational and cognitive biases. In this section we find areas of opportunity as diverse and everyday as inducing responses or tiring the interviewee. In the first case, we must verify that basic principles are followed when applying the questionnaires, such as reading randomly precoded response lists to avoid first response bias or reading the questions as they appear in the survey instruments so as not to alter the sense of the answers. In this section we can also mention avoiding suggesting the solution with the question. Secondly, the very long interviews that end up tiring the people who collaborate with us and who will seek to free themselves as soon as possible by answering without further reflection should be supplemented as far as possible with the fewest possible questions.
  4. Investigate blindly. Although it might seem obvious, quantitative studies should start, whenever there is antecedents, from desk research. In this way, you can have a more complete overview of the segment to be searched so that you can answer our interviews. On some occasions, samples are designed that are not careful to define a detailed profile of who can provide useful information for our purposes. An example of this type of error would be that of a case in which we need to know the uses and customs of active practitioners of a sport discipline and we are not clear on the definition of «active practitioner.» In these types of errors we include being little selective or permissive with those who answer our questions. Some possible solutions to alleviate this problem are to introduce decoy questions to ensure that the respondent does know the subject of which they share their experience or use the instinct and common sense of the interviewer to detect enthusiastic people, but ignorant of what we want to deepen.

Avoiding non-sampling errors is one of the basic skills that every market analyst must include in his repertoire from the beginning of his professional development. We know that the difference between an investment in research and an onerous expense depends on adherence to this guideline. The analysts and surveyors who work for Acertiva fulfill our mission to be the most trusted market research company. Do you have a project on the doorstep? Write to us today so you can share your needs with us and we can tell you how we can solve them for you.