Copied from here, where I'm getting no replies Hi All, We have a GACE (Google Analytics Content Experiment) with 2 variations (original and new), and the "new" variation is showing > 100% conversion rate after 344 experiment visits, which is clearly not possible (our conversion rate is around 18% overall). In addition, the "original" variation also shows a surprisingly high conversion rate (> 23% in all 3 experiments, when 18% would have been more likely). We are using Google Analytics since November of 2006. In the conversions, ecommerce, overview area I can see that conversions are tracking 100% correctly. GA is something that we use on a daily basis, and I'm fairly certain that the transaction tracking is dead on. We have been using GACE since October of 2012, with close to 100 experiments performed. Our method has always been the same - showing the same page with a URL param to indicate which of 2 versions we show. The GACE code is shown on both variations of the page - however, showing it on both versions of the page has never been a problem. Details: Objective for this experiment: Transactions Percentage of traffic to experiment: 100% Minimum time to run: 1 week confidence threshold: 95% Original: www.visiondirect.co.uk/?xFrontPage4=N Variation: www.visiondirect.co.uk/?xFrontPage4=Y Original: 256 experiment visits, 74 transactions, 28.91 ecommerce conversion rate Variation: 88 experiement visits, 90 transactions, 102.27% conversion rate, 254% compared to original, 100% probability of outperforming original. Using Google Tag Manager, so the Google Analytics Tracking code is not found, but the analytics code IS found. IMPORTANT: This is now the 3rd attempt to test the same changes that is showing this problem. The first time, we started the experiment without including the analytics code (whoops), so when we added the code and found skewed results, we were not too surprised. We stopped the test, and copied it, and had the same problem with the second test. "Ah, they still have the cookies from the first two test" (same experiment ID), "so I'll start a totally new experiment from scratch". However, even with new Experiment ID and Experiment Key, I am having the above problem. Experiment ID:cr4ZE_XkQlyX4r5O55xmtQ Experiment Key:212272-76 Any ideas as to what is going wrong? Thanks, Ed Eichman Cambrils, Spain