To me, there are two parts of a successful test: 1. Statistically significant increase in conversions 2. A reasonable understanding why behavior changed In th

To me, there are two parts of a successful test: 1. Statistically significant increase in conversions 2. A reasonable understanding why behavior changed In the case of the larger button converting less, if form progression were recording and found that fewer users even attempted the form because of a perceived spammy CTA, then the test is a "success". Without the data, you don't know why, so you can't really apply your findings authoritatively. Running tests to maximize revenue is one thing, but the smart way is to understand user desire and behavior on your site, and create an experience conducive to that.