In the face of that immaturity we can try all kinds of things. Personally, I do anarchist development and hope for the best. Others try cargo cult project management. Still others move to highly constrained development methodologies.
It also implies at least, that we can collect requirements in some adequate way, which is really a circular argument. Requirements gathering is even less mature than other parts of software development.
A/B testing helps the process of testing assumptions in a data driven way.
As for surveys, KPIs, A/B testing etc. Those are neither standardized, proven to be valuable in a rigorous way, nor widely used. All signs of an immature industry. Further, they can only be used in software development processes that allow for iterative deployment.
The harder question I suppose is whether it was successful and we've got/maintained a quality codebase.
Thats an example of a successful pregnancy project.