Maybe I missed something, but it seems like measuring the winner based on average time of the first couple of samples is significantly flawed. This is equivalen

Maybe I missed something, but it seems like measuring the winner based on average time of the first couple of samples is significantly flawed. This is equivalent to me putting you in front of a keyboard with a normal layout, and then a layout where the keys are scrambled and asking me to type a sentence, and then measuring success by average WPM on all attempts. ... ofCOURSE i'm going to be slower on the different layouts regardless on my first, second, third, maybe even hundredth use. But 2 months down the road the benefits may start rolling in. It seems to be that a better way to measure results would be to create an Android only app (iOS doesn't support swapping primary dialer). and getting participants to use your dialer for a couple months, once you notice a new baseline per user compare results between the various treatments. Tl;dr unless I missed something! these results aren't going to prove anything, other than familiarity rules. All that aside -- I LOVE this! In the spirit of experimenting and trying to improve on a very old concept, and creating an app to whip up some quick results, very, very cool!