"Deliver first, improve later" This is the approach behind “minimal viable product”. It is also valid for data science solutions.