Blog

Questionnaire Design: Response Scales - Part II

Part II: Choosing a Response Scale - Granularity

My last installment on the discussion of response scales addressed why we should use them. So, once you’ve decided to use a scale, you’re faced with the decision of what kind of scale to use, and what it should look like. There are many ways to present response scales, and each one has its particular strengths and purpose.

More points means greater sensitivity

As with everything in research, I always go back to the objectives: what is it that you want to achieve, and how do you intend to use the results? This will drive every aspect of questionnaire design, including response scale use. For example, let’s say you want to understand customer satisfaction so that you can track changes between specific time periods. Let’s say these time periods are fairly short, and as a result have somewhat less wave-to-wave fluctuation. You may therefore want the response scale to be more sensitive in order to be able to pick up on the smallest of changes, so this will mean using a scale that is more fine-tuned, that is, one with more points on it. This would enable you to act nimbly on any change that was recorded from wave to wave.

Even or odd?

So when you’re deciding on how many points to include on your scale, another decision you need to make is whether it will be an odd or even number of points. An odd number will provide a mid-point, which may or may not be desirable, again, depending on your objectives, and how you intend to use the information, as well as whether you want to avoid a neutral response. For example, perhaps you are measuring satisfaction, and you really want to know whether your customers are satisfied – you want to force the issue and get them to choose a positive or negative rating, without any sitting on the fence. You may decide on a scale that includes the following points:

Very dissatisfied

Somewhat dissatisfied

Somewhat satisfied

Very satisfied

By not offering some point between dissatisfied and satisfied, you force the respondent to commit to one side or the other, presumably in order to know whether/how much you need to improve your offering.

However, let’s say you’re asking about some hypothetical future behaviour, such as intent to purchase. It may be possible that the respondent truly does not know whether they will carry out this behaviour, because they require more information (costs, benefits, their future needs/situation, etc.) in order to make this decision. Similarly, if you are asking opinions on topics that the respondent has never before encountered, it is possible that they have not yet formed any opinion on the subject. In such cases a mid-point will offer a position where they can respond without committing to a positive or negative answer. Alternatively, if an even-numbered scale were offered in addition to a separate “Don’t know” category, this could achieve the same result. Either way, we need to be sensitive to situations where forcing an answer one way or the other may be inappropriate.

So in the end, when deciding on number of points in your scale, think ahead to the analysis and how you want to present results, think about the nature of the measurement you’re taking, and think of the respondent experience in answering your question.

But perhaps you may not want to be bothered by minute fluctuations, in cases where they may be distracting, less meaningful, or in some way not worthwhile to act upon. Perhaps the measurement normally fluctuates a great deal, and you need to obtain a sense of the bigger picture, where fluctuations go beyond the norm. In these cases, a smaller number of scale points might be in order. Again, there may be other considerations that weigh into your choice, and these are just some examples of the factors affecting your decision.

Leave a Reply