The name, A / B testing, speaks for itself: it experiments with graphics and text. Upload multiple creative materials, select a percentage (%) of users for each variation, and wait for the results. With a lot of downloads, 10-14 days are enough for the test. Sometimes, however, the result of an experiment becomes obvious during the first week. To get a more accurate picture, do not limit yourself to A / B, but also B / A, A / B / B, and other tests. Using these methods, you can check the reliability of the first results. Try not to change anything in the application during the experiment since that might affect the indicators, and you won't be able to figure out if the experimental features or other factors were responsible for that.