Web surveys are frequently faster than other modes, less expensive, and with modern software they can accommodate even the most complex questionnaires. However, problems abound, including limited Internet penetration, no email address analogue to the national telephone number sampling frame, and reasons to suspect that people may respond differently on the Web than they do on the telephone. This last issue-potential mode differences between Web and telephone-is the main focus of this paper. The survey data for this research are drawn from a monthly customer satisfaction survey for an energy distribution company. Representatives from a sample of large business customers are asked to rate the company on a number of satisfaction criteria. Each month a new sample is drawn and an invitation to complete a Web survey is mailed to representatives at each company selected. Non-responders are followed up by telephone. On average, each month, we achieve a 50% response rate with roughly 40% of the respondents reporting by Web and the remainder by telephone. Our analysis has found significant differences across modes for a number of items, many of which are consistent with mode effect hypotheses that have been posited between self-, and interviewer-, administered modes. Self-selection of mode has partially biased our results, and in the spring of 2005 we will test these differences by randomly assigning sample members to either Web or telephone treatments. This research provides valuable insight into the impact of mode on response to a number of commonly asked customer satisfaction questions. The research expands our knowledge about respondent preference for survey mode and assists efforts to improve participation rates for surveys that employ a multi-mode design. The paper also provides some practical experience on the problems associated with transitioning respondents from one mode (telephone) to another (Web).