Rate our products / company on a scale of 1 to 5 – probably all of us met with this type of question: in shopping satisfaction surveys, customer service department reviews, satisfaction with a doctor’s visit, etc. Sometimes, however, the company wants to ask for more than one product or point of contact with customer. In such situations, the matrix question will be perfect. Here is an example of such a question:

How is the average in the matrix question calculated with a numerical scale?

To analyze the results obtained on the basis of such a question, we need to calculate the average.

If the answer scale is based on statements / words, they have default weights (1, 2, 3 … in order). In this case, the average is calculated from the indexes of these answers, e.g. 1 * x answers + 2 * x answers + 3 * x answers / number of answers = average.

In another case, when the scale of answers is based on numbers, as in the example above (scale from 5 to 1), the average is calculated from the values ​​of these answers.

Let’s analyze two cases with a numerical scale:

Case 1:

Scale: 5,4,3,2,1 + answer „I don’t know”, which in the survey settings is marked as a neutral answer (on how to set it, read in this article). Example:

In this case, the „I don’t know” answer does not affect how the average is calculated. For the scale: 5,4,3,2,1, 'I don’t know', the average will be calculated from these numerical values.

We calculate it this way:

5x number of answers + 4x number of answers + 3x number of answers + 2x number of answers + 1x number of answers / sum of answers (we do not include „I don’t know” because it is marked as a neutral scale).

Case 2

However, if at least one answer does not consist only of numbers and we don’t mark it as a neutral scale, then the average is calculated from the indexes of these answers. In this case, when calculating the average, we include the answer „I don’t know”.

We calculate the whole thing this way:

index1 x number of replies + index2 x number of replies + index3 x number of replies + index4 x number of replies + index5 x number of replies + index6 („I don’t know”) x number of replies / sum of all answers.


In the case of answers in the form of numerical values, mark any additional „I don’t know” answers as a neutral scale. Only then can the average be calculated correctly.

Did this answer your question?