How can we change survey response scales on existing surveys without data loss? | XM Community
Skip to main content

We are considering switching from a standard 7 point scale to a 5 point scale to simplify the user experience across all questions on all surveys, so that on all attitude/opinion questions we use 5 points (except NPS). Most of these questions are single select multiple choice questions.
Since we are pushing data to our data warehouse and pulling it into Tableau for reporting, we need to keep the same project and ideally also preserve the same questions from last year so that we can maintain the data in the same field names for year over year trending. .
We'd like to either hide specific points on a scale, or replace the scale entirely with a new shorter version, without losing the data associated with responses that use the old format.
Has anyone tried this before and if so, what worked/did not work?

As long as you don't delete scale point, and just hide them, your old data integrity should remain. If any of them overlap, you can consider keeping some of the old scale points visible and hide others, add new ones. Or you can choose to hide all and add all new ones. Probably depends also if you are recoding values, since these recodes will be different with the new scale.


I don't recommend modifying the original question, since you would have to relabel the numeric values of the scale and that would effect previous years' data (if all responses are in the same qualtrics project). This will lead to confusion and potential data reliability/integrity issues when it comes to analysis.
Instead, you could create a new questions for the 5 point scale, and simply hide the old 7 point versions/set them to not display for anyone (maybe put them all in one block that's not a part of the survey flow, or leave them as is and set display logic to true - false = false).
Assuming this scale is bipolar, you can create a new variable that recodes the 7 point versions to the 5 point format when running analyses with the trend data, if necessary, by following this logic with the response options: 1 = 1; 2 & 3 = 2; 4 = 3 (the midpoint); 5 & 6 = 4; 7 = 5. Then, you could create a new variable that pulls all responses for the 5 point variables across all years.


Leave a Reply