Suppose a brand wants to know whether running a display ad campaign on a news website increases sales. To that end, they plan to run an A/B test in which they will randomly split the traffic into two equal groups---one which sees the ad and the other which sees a filler ad (unrelated to the brand). Given that the average conversion rate (purchase per view or impression) is 0.2% without the ad campaign, how many impressions are required per group to detect a 3% improvement in conversion rate at the 5% significance level with 80% power?