While the confidence level of an app is the primary lever in adjusting how quickly our algorithm is able to run through a test, the minimum detectable lift (MDL) comes into play and tells how the algorithm should make decisions when assets perform very similarly toward the end of a test.
The MDL refers to the minimum relative lift above the Control you're looking for to deem it a successful test. Typically speaking, within our ASO test environment, if we found an asset that's just 1% better than the Control (e.g. Control at 20% conversion rate and a variant at 20.2%), it is not convincing enough of a winning margin to replace the Control in the live app store.
Furthermore, as you set a lower MDL, if 2 assets continue to perform neck and neck, it may require exponentially more traffic to find a winner at a certain confidence level before the algorithm closes the test and moves on. You may then have to make a manual decision to select a winner as our algorithm won't move on automatically until a winner is selected.
How to Edit the MDL
We recommend an MDL range of 5-10% (the default is set to 8%). You can edit this as needed in the "Routing & Options" page:
