As response rates continue to decline it is crucial to research areas that examine the conditions under which survey nonresponse results in nonresponse bias. Although there is a well-established body of literature that discuss methods to increase responses rates in web surveys, less is known about the effect of declining response rates and the link to nonresponse bias. The Snapshot survey, a representative survey of the membership of an integrated healthcare system in Washington state, included a randomized experiment testing various permutations of operational modes to measure the effects of nonresponse to a web-based survey. The use of administrative data from medical records allowed for the comparison of response rates, demographic characteristics, and health information of responders and non-responders. The experiment provided the opportunity to investigate multiple facets of nonresponse. Are the operational designs significantly different in terms of their resulting response rates, the demographics of the respondents and non-respondents and in the overall results? This paper will discuss the descriptive findings of the randomized experiments and discuss implications for future web-based health surveys.