Testing is the best antidote to decisions by assumption. Whether we work from assumptions because they're our own unquestioned beliefs or because they're the opinion of people with position or power, they are never as effective as question-based testing. Always be testing something. Always have some kind of tests going on for both improvements and innovations of deliverables and processes.
Test new variations. Always test two or more approaches different from the norm in order to drive richer learning.
Do short tests. Do only as much testing as it takes to establish a pattern under fairly consistent conditions. The shorter the testing cycles, the more options can test and the more tests you can do with the same resources.
The principle at work here is simple: growth is learning and learning is experimenting.