Surveys mainly use questions in which it is allowed to answer only through a closed series of alternatives. The choice of labels for these closed alternatives is an important decision. Depending on this choice, different results can be found. This paper focuses on the impact of using low versus high frequencies or durations scales. The novelty is that it studies panellists of an online panel oriented towards marketing surveys. Also, it uses data from countries little studied before: Spain, Mexico and Colombia. Using a split-ballot experimental design, it shows that significant differences in answers are obtained depending on the scale used. In order to determine which scale gives results closer to the reality, the correlation with an external variable is used; the higher this correlation, the better the scale. In practice, this information can and should be used to select the best scale for a survey.