There are some common mistakes I see too often in training feedback questionnaires (as well as in any type of questionnaire I’m asked to complete). I find them annoying and they make me answer less accurately. To help you create better questionnaires I’ve compiled the top 5 mistakes in my experience. Adhering to all of the points below will get you a much fairer result as well making your respondents happier overall. Have a read through and try to remember when creating your next feedback form! Was there something you didn’t agree with? Leave me a comment!

1. Include the middle ground

When working with Likert-scales you have the option to have either an even or odd number of possible answers. Whichever you decide will impact the results you get, and before deciding on your scale you need to understand how. As an example, compare these two tables:

 Even scaleOdd scale
Fully agree20%20%
Somewhat agree30%20%
Neither agree nor disagree-20%
Somewhat disagree30%20%
Fully disagree20%20%

The people choosing the middle option 3 in an odd scale are forced to make a decision with an even scale. They will lean either way and logically their choices should follow those of the other respondents, giving the same number of answers towards both directions. With an even scale, you could say that 50% agree with the statement whereas with an odd scale you could say that 40% agree with the statement. Which is more correct? Rather than answering this question right off I’d propose answering it giving all the statistics when presenting the results: 40% agreed, 40% disagreed and 20% did neither.

Though both an even and odd scale work, in my experience you need to be careful when selecting an even scale. Respondents may be too lazy to correctly decide what they want to answer when they are forced to. Giving them the option to say that they have no opinion will in general make both you and the respondent more satisfied.

2. Define the scale

Did you ever consider that yours and the respondent’s view of a scale are not the same? Assume the respondent wants to answer “very good” to the following question:

On a scale 1-10, with 10 being the highest, how did you find the training material?

Where would you put “very good”?

Are you sure other respondents would do the same?

Depending on how the respondent defines “very good” the answer might be anything on that scale. Perhaps the respondent sees 1 as bad and 10 as very good, perhaps he/she sees 1 as good and 10 as excellent? When you later evaluate the response, how do you know the respondent’s definition of the scale?

To avoid these problems you always need to have the scale defined. At the very least define the top, bottom and middle points. Preferably every point in the scale should be defined, reducing the number of definition-errors. Also aim to use fairly simple words for the definitions, in order to overcome language problems.

3. Don’t ask the impossible

Even at top-name consulting agencies I’ve come across surveys asking for a Yes/No-answer using a graded scale. This surprises me as trying to answer such a question, while maintaining logic, is impossible. Consider the following question:

Would you recommend this training to your colleague?

1. Fully agree
2. Somewhat agree
3. Neither agree/disagree
4. Somewhat disagree
5. Fully disagree
– No answer

The question in this case is asking for a Yes or No answer while the only answering possibilities are how much you agree. This question either needs to have other answering options (Yes/No/Maybe) or to be rephrased into a statement, like this:

I would recommend this training to a colleague

4. Have an option to opt out

Even though you’d like a 100% response rate on all questions, you can’t force the respondent to answer questions he/she doesn’t have a clue about. The difference compared to the middle ground option discussed in the beginning is that not all questions may be applicable, even if the respondent would like to answer. Consider the below question:

How would you rate day 2 of the training?

5. Excellent
4. Good
3. Ok
2. Bad
1. Awful

What if the respondent was ill and couldn’t take part on the second day, how should he/she respond? Ensure that you always include an option saying No answer, Not applicable or something similar. This way you will have less frustrated respondents, and more correct results. Also remember that when presenting the results you need to say specify whether your percentage is of the total number of respondents (including the n/a-answers), or of this question’s number of respondents (excluding the n/a-answers).

5. Avoid negations

Consider the following statement:

I’m not unhappy with the seating.

If you liked the seating, should you answer that you agree or disagree with this statement? Most people will have to think some before answering this question, something that will both confuse the respondent and may cause incorrect answers. In this example, if you agree with the statement then you are saying that you like the seating. Even if the respondent may spend enough time to correctly answer this question, will the reviewer correctly identify the answers?

A much better way to ask the very same question would be:

I’m happy with the seating.

By removing both the negating not and the negative unhappy you make it much easier to work with this question.