What does small STDEV tell you about survey data In measuring Customer Satisfaction using surveys, consider you have a 15 question survey questionnaire and each question could have 1 of 11 responses (0 to 10, with zero being Extremely Dissatisfied and Ten being Extremely Satisfied).
With the growing condition of "Survey Fatigue" that is born of the fact that consumers or inundated with survey request, I wish to know if concerns over the quality of these data sets can be determined mathematically using STDEV.  One might expect a certain amount of variance among questions that ask for customer perception responses to very fluid situations.  However, we are seeing increasing low STDEV of the datasets, leading to concerns that results are being "pencil whipped" rather than thoughtfully considered with more surveys coming in all top box responses.  I'm just not sure that STDEV alone can make this determination.  Any help with this is appreciated. Best, Robert
 A: As a market researcher, Std Dev is not a good measure of engagement. One thing that I have used it for is to determine straight lining across batteries of like questions.
At the end of the day, how you expect a sample to respond to your survey is not necessarily how they will respond to a survey.
What you need to do is take a hard look at your survey and determine whether the questions you are asking are at such a nuanced level that the survey taker may not be able to tell the difference between one attribute vs. another. This leads to confusion and frustration which are going to more negatively impact the quality of your results than the frequency in which people are being asked to take surveys.
I recommend you reach out to ESOMAR as they have some good best practice white papers on survey design. If you are a supplier side researcher, you will need to push back on your clients demands. If you are a client side researcher, you will need to understand that the nuance in which you view your product or service is much more granular than what your customers see and develop less hyper specific attributes and questions they cannot accurately answer. 
