I am often inclined to put the implementation questions first, ie, "Can anything be changed?" Should the implementation question not accompany the whole process from its very beginning to its very end?

Churchman (1979, p. 21) as cited in: Interfaces (1982) Vol 12, p. 12