A lot of websites use web based exit surveys in order to measure the satisfaction of their visitors. The related literature includes user satisfaction studies of library, health related and other more commercial-oriented websites. Usually the response rate to these web surveys is very low. This low response rate raises questions about the quality of the data collected by the web survey. In this paper I try to provide some answers to the question if the sample is representative of the total population of website visitors.
Methods & Data:
The findings presented in this paper are based on the analysis of data collected by the Greek voting advice application HelpMeVote and the analysis of data collected by the corresponding exit survey. The setting of HelpMeVote is perfect for the comparison between the total set of visitors and the subset of people who have responded to the web survey. The only reason someone visits a voting advice application is to answer to a series of questions in order to get his/her proximity with the political parties. Before giving the output I ask users to fill-in a form with their personal information. Although it is not mandatory (users can click “continue” and move on to the output without answering the questions of the form) the vast majority responds to these questions. As a result I have the distribution of the population and I can compare it with the distribution of the sample.
Results:
Logistic regression analysis shows that the probability to respond to the website exit survey is larger for male, younger and more educated visitors, but the most significant predictor of responding to the web survey is the level of satisfaction. As a result, satisfied users are over-represented and unsatisfied users are under-represented in the sample.
Added Value:
The findings of this paper demonstrate that visitor satisfaction estimated by a web-based exit survey will be higher than the visitor satisfaction we would measure if non-respondents would be included in the calculation.