Not all scales in market research need be of the Likert Scale, Semantic Differential Scale, or Constant Sum variety. They don’t have to involve Bayesian theory like the extensions of Maximum Difference Scaling doin their process. In fact, sometimes the simplest approach truly is the best.

The multiple response format is a common question type found in market research surveys. Essentially you are giving the respondent the option to select any or of the options you provide. For example, the following two questions are used to create the Information Seeking (iSeek) scale. Each asks the respondent to select any or all of the activities they participate in.

In 2014, which of the following have you participated in or attended? (Select all that apply)

  • A classroom (out of office) training session
  • An informal learning session at work
  • A CD or DVD-based training tool
  • A self-paced online training session
  • A live instructor-led online training session
  • A formal training session at work
  • Other (please specify)
  • None

In 2014, which of the following methods have you used to obtain information to increase your working knowledge? (Select all that apply)

  • Attending a professional seminar, luncheon or technical conference
  • Downloading a white paper
  • Joining an online community
  • Researching a topic online
  • Reading and/or contributing to a blog
  • Posting to or following someone on Twitter or LinkedIn
  • Web seminars
  • Using a mobile device to download applications or view content

Collectively there are 15 functional categories that can be selected. It is important to ensure the categories used are randomized, with the exception of the “Other” and “None” categories. Each category is a dichotomous (selected/not selected) question in its own right. This allows the researcher the option of assigning a zero for not selected and a one for selected. This means you can sum the categories and this forms the basis for our scale. Respondents can be divided into equal groups based on the summated variable.

If you are feeling lucky you can leverage a clustering solution with the dichotomous variables. This will yield a segmentation based on response patterns. Latent Class Analysis is well-suited for this exercise. You can also ask a third question which tasks the respondent with ranking the options they selected. This provides a measure of importance which you do not have just the multiple response question.