There are moments in online panel operation when technical, financial, or organisational constraints impose a change in panel functioning rules, resulting in increased risk of attrition. On the occasion of panel functioning change in ELIPSS, a French non- commercial online panel, we explored information predictive of panel survival when membership became less attractive to panellists. Available information was collected at different stages of individual participation: recruitment, profiling, and survey response behavior (paradata). We also devised an experiment aimed at testing differences in effectiveness between three types of incentives on panel survival: repetitive financial incentives (two incentives separated by a four-month period), single financial incentive, and single financial incentive coupled with a gift. A hierarchical logistic regression model made it possible to distinguish indices derived from survey response behavior and from questions pertaining to panellists’ initial motivation to join the panel as significant predictors of panel survival. Our findings can be useful for panel monitoring in the context of important changes in its functioning. They outline that long-lasting attitudes towards panel participation have much more impact on panel survival than the amount of incentives received.