The real cost of building everything
Every feature in version one is a feature you must design, build, test, host, support and eventually migrate. The cheapest feature is the one you did not need.
We ask a simple question of each requirement: if this were missing on launch day, would a paying customer walk away? If the answer is no, it belongs in a later release.
What a good MVP still includes
Cutting scope is not an excuse for a fragile system. Authentication, payments, data integrity and deployment automation stay in scope, because retrofitting them is far more expensive than doing them once.
The parts that shrink are breadth of features, admin conveniences and edge-case handling that can be done manually while volumes are low.
When a full build is the right call
If you are replacing a system that people already depend on, an MVP is the wrong shape. Migration projects need parity before switchover, so scope is set by the legacy system, not by ambition.
The same applies in regulated industries where partial compliance is not an option.