It depends. You cannot have a "good enough" with rapid iterations approach for a mission critical project. e.g. A software to control chemotherapy or a missile

It depends. You cannot have a "good enough" with rapid iterations approach for a mission critical project. e.g. A software to control chemotherapy or a missile controller program However for consumer facing products or in some cases for enterprise products, you can start with a minimal working version and then improvise. Personally, i prefer A because the real feedback comes in from your consumers and you will find lot of your assumptions going wrong