I believe in the modern approach of user testing : submitting your design to at least five users matching the profile of those expected to use it is a highly-valuable asset for a user-friendly design.

To do so, I go through four steps : First, I prepare user-testings with high-fidelity prototypes described with assumptions. Second, I prepare a test guide and make sure a list of users is contacted and ready for testing. Third, I go through user-tests helped by another designer that will take notes. And fourth, I regroup all the data gathered and confront its results to the original assumptions, then share this synthesis. 

User testing has two costs : designer time, and available users. As of now, I chose to focus user-testing on complex features, on those with a high product-value, or on those that returned negative feedback. 

Data treatment can go (roughly) two ways : discardable, or atomic research. 

Discardable data treatment is time efficient : I use the data collected for the tested flow only, then I store it and don't use it anymore. 

Atomic research data treatment is time consuming, but has a higher product value : I store the data using various tags, so it can be crossed-used and benefit the whole product on a longer period.