In collaboration with a large charitable organization, we send out upgrade requests to regular donors via postal mail and vary the campaign progress illustrated by a progress bar. In contrast to recent studies, we differentiate between the effects of extreme goal gradients (very low or very high) and effects of a moderate goal gradient. We randomly assign receivers to either one of the three goal progress treatments or to a control group. In the control group letters, there is no explicit information on the existence of a campaign goal or on behavior of other donors. In an additional treatment variation, we include an information (and a bar graph) unrelated to campaign goals which serves as a control for the letter design. Donors, whose e-mail-address is available to the charity, receive an additional e-mail with the same content two weeks after the postal mailing. Between treatments, we compare upgrade incidences and upgrade sizes.