Deep diving into a vision allows me to discover the problem by mapping out user flows and looking at data.
Story lining the desired flow of the user to understand the thinkings of what, why, who and how.
Getting a feel for the story arc through low-fi prototypes. These could be static or interactive.
Validated hypotheses with qual and quant data, tested remotely, contextually and post launch.
Taking the findings to improve the UX, based on real data, repeating the build-test-learn cycle.
Post launch testing is valuable, setting up A/B or MVT's to squeeze the last few conversion percents.