Surveys enable researchers and designers to gather information about website usability from users around the world quickly and efficiently. They can even replace the need for direct interviews or observational user testing in certain situations. But understanding the usability testing scenarios for which online surveys work best is equally as important as knowing how to devise them.
Online surveys are not a complete substitute for other forms of usability testing. Nothing can replace the interactivity of an interview or watching a live user test with a working prototype. Similarly, what users say they do on the Internet does not have the statistical reliability of tracking data. What online surveys can accomplish is information gathering about your target audience: their background and demographics, their personality, their usage patterns and preferences, etc. It is also useful in understanding how they feel about your existing site and its competitors. This is a prerequisite for developing other early stage tools, such as personas, workflows, use cases, and requirements. Online surveys can also be used to quantify the success of a project after its completion, in order to validate that users can achieve common tasks or to measure whether or not an acceptable level of ease-of-use and approachability has been achieved.
Once an appropriate application for a survey is decided upon along with a clear idea of the information you are trying to gather, it is then important to craft the survey to the task at hand. There is a fundamental opposition in online surveys between the sponsor and the participant: the sponsor is trying to gather the most accurate and complete information from a user who is trying to complete the survey as fast as possible. When possible, thoughtful incentives and careful participant selection can best help preempt these concerns. In any case, you must design the survey for the participant by making the survey easy to use and not at all intimidating in difficulty or length. Like questions should be grouped in a way that is meaningful to the user, appropriate page controls should be chosen (checkboxes vs. radio buttons vs. text fields), and user frustration should be avoided at all costs. It is also important to communicate to the user their progress in order to avoid abandonment, a common problem in online surveys.
When the survey has been completed by the appropriate number of participants, the results should be analyzed to first understand how valid and useful they are. Looking at abandonment rates, junk or hurried responses (including empty responses or, for example, participants who consistently chose the same response in a multiple choice scenario), and inconsistencies can help to tell early on whether or not the survey results are accurate. Once enough valid results have been gathered, both statistical and anecdotal analysis can begin in an attempt to understand your users and find the context in which to design.