One of the biggest problems with customer surveys today is the use of survey processes which systematically exclude certain customers from participating. For example, e-mail surveys can only be taken by customers with valid e-mail addresses, and end-of-call phone surveys can only be taken by customers who don’t hang up before the survey.
It might not matter that these people are excluded from the survey, but in many cases it matters a lot. Upset customers often hang up before taking a survey (they’re just glad the call is over), making the survey results look better--sometimes far better--than reality. Some survey processes are also vulnerable to manipulation by employees eager to improve their scores.
The best way to get a handle on whether the survey process is biased is an occasional “reality check” survey using a completely different method. There should also be an active audit and appeal process, and you should be on the lookout for anything strange about the survey. If a customer takes the time to report unusual behavior (for example, a salesperson offers an incentive to customers who give positive surveys), that’s a giant red flag that the survey is probably being actively manipulated.