Because of its interactive character the web may promote more accurate survey data than do other, more static, modes. One way that interactivity can increase data quality is by providing feedback to respondents about their answers. For example, if questions require that multiple answers add-up to a fixed total, e.g. 24 hours or 100 percent, then inviting respondents to revise answers that do not sum to this figure could increase the number of appropriate sums. We report a study that compares respondents' estimates of percent time spent on each of 10 randomly ordered categories of internet usage when they did and did not receive feedback. 1788 web respondents were randomly assigned to one of three conditions: No feedback, Server-side feedback indicating that the submitted responses did not tally to 100%, and Client-side feedback in which a running tally was incremented as each response was entered prior to submission of all responses. Clientside feedback was also accompanied by a server-side message if necessary. The server-side feedback was displayed once, after which the system accepted all answers. The results show a clear advantage for Client-side feedback. More answers tallied to 100% in the Client-side than Server-side condition with the smallest number in the No Feedback condition, and responses were faster in the Client-side than Server-side condition suggesting that the running tally promoted revision before respondents submitted their answers. Client-side feedback seemed to promote more thoughtful revisions as well: when the Server-side feedback indicated the sum did not equal 100%, respondents who had not been given the running tally changed answers to the first category most often - regardless of what it was - but respondents who were presented the running tally changed answers equally often for all categories. We conclude by discussing the promise of interactive features in general for survey data quality.