Simply put — “seeing is believing.” It isn't until working code is written that the project team knows the architectural vision works. Most modern processes suggest that a working, end-to-end "skeleton" of the architecture be built to prove that all aspects of it work. This can make choosing the right feature set to be developed initially to be a challenge, as it is rare that a “simple set” of features will exercise all of the requirements of an architecture end to end. Perfect is the enemy of good enough! For example, in the case of a Data Warehouse solution, the minimum requirement might be that the major legacy data sources be accessed, the ETL strategy works, the database regression testing strategy works, and that the reporting tools can access your DW.