You're testing the effectiveness of a new ongoing drug counseling program (vs. another, already established one). You randomly assign people to either the new program or the established one. You find that, over the course of the study (1 year), significantly more people dropped out of the new one. You found that the average level of 'success' among those in the new program was higher than the success of those in the old program. What confound might that create, and why