Some things in life just don’t mix. Oil and water. Gasoline and an open flame. Well in the world of consumer and B2B marketing research, the question whether numbers should be mixed with text descriptors remains when it comes to creating a scale on your questionnaire. This is a house divided as one camp says that numerical scales do not need descriptors while the other camp says do away with numbers all together and go with verbal scales. Let’s take a closer look.
On one end we could ask our respondents to rate their level of customer satisfaction with a simple numerical scale perhaps one to ten. In western, particularly the United States, we have a bias toward thinking that lower valued scores are less positive than higher valued scores, with the possible exception of ranking problems. It is advisable to state, perhaps in the question, what the anchor points represent, e.g. one (1) is not satisfied and ten (10) is very satisfied.
An alternative is to do away with numbers (as far as the respondent is concerned) and provide text responses. There is some evidence in the literature that respondents prefer rating scales with verbal labels over strictly numerical labels, and that they offer the researcher greater reliability and validity. The challenge is determining the appropriate slate of verbal labels to match the number of points your scale requires. For example:
Is there a compromise position that draws upon the simplicity of numerical scales, but offers greater reliability and validity? I say yes there is! The example below uses an 11-point scale similar to that of the Net Promoter Score question. It also combines a verbal scale. Although I have no empirical proof, it seems logical that a question displayed in this format speaks to both the numerically and verbally inclined.
If you currently have standardized scales for your research purposes, I suggest conducting an A/B test to see if changing the format to one that combines both verbal and numerical alters the average scores. If there is no difference and space permits on your questionnaire design then try a format that speaks to both sides.