All related (6)
Puja Hait
Product Leader, GoogleSeptember 9

I would say there is no substitute to real user data.  

User research is table-stakes. But in my experience, not always representative of "actual" usage, so don't overindex on specifics. Rather validate if the problem statement is indeed important for the user segments. Because if it is, then they might be more patient with your iterations towards solving their needs. Else they are very likely to abandon very early on and never return. A good way to test is whether they are willing to pay for the solution.

Paper mocks UXR is helpful once you have narrowed themes and want to develop MVP scope. 

Then go build MVP and ship to a small user segment to learn and iterate. This step is the journey. 

Scale up only when you see hockey sticks for atleast one feature. Also retention loops should be naturally showing up by then.