Scope that is too big to finish or too small to count

Two opposite scope problems account for many rejections:

• Too big: "A complete hospital management system with AI diagnosis, billing, pharmacy and a mobile app." No team finishes this well in two semesters, and the committee knows it.
• Too small: a basic create, read, update and delete (CRUD) app with login and a few forms. It may work, but it does not show final year level problem solving.

The fix for both is the same: one clear problem, a must list that fits the time, and at least one part that is genuinely challenging, such as a non-trivial algorithm, real-time behaviour, a model trained and evaluated on real data, offline sync, or a hardware integration.