I know I'm going completely against most training-industry practice in saying this, but it's the truth. Likert-like scales create poor data on smile sheets.
If you're using questions on your smile sheets with answer choices such as:
- Strongly Agree
- Neither Agree Nor Disagree
- Strongly Disagree
You're getting data that isn't that useful. Such questions will create data that your stakeholders--and you too--won't be able to decipher very well. What does it mean if we average a 4.2 rating? It may sound good, but it doesn't give your learners, your stakeholders, or your team much information to decide what to do.
Moreover, let's remember that our learners are making decisions with every smile-sheet question they answer. It's a lot tougher to decide between "Strongly Agree" and "Agree" than between two more-concrete answer choices.
Sharon Shrock and Bill Coscarelli, authors of the classic text, now in its third edition, Criterion-Referenced Test Development, offer the following wisdom: On using Likert-type Descriptive Scales (of the kind that use response words such as “Agree,” “Strongly Agree,” etc.):
“…the resulting scale is deficient in that the [response words] are open to many interpretations.” (p. 188)
So why do so many surveys use Likert-like scales? Answer: It's easy, it's tradition, and surveys have psychometric advantages often because they are repeating the same concepts in multiple items and they are looking to compare one category to another category of response.
Smile sheets are different. On our smile sheets, we want the learners to be able to make good decisions, and we want to send clear messages about what they have decided. Anything that fuzzes that up, hurts the validity of the smile-sheet data.