An A/B test is run on a website where users can visit multiple times. Each visit is treated as an independent observation. What statistical problem does this introduce?