You want to compare multiple distributions on the same axes (overlapping KDEs are clearer than overlapping histograms) - You want to emphasize the smooth shape of the distribution - You have enough data points (at least 50-100) for a reliable density estimate