Cervelo's presentation is very good, but they do not mention that stack and reach are dependent on each other. A frame with a higher stack is effectively "longer" than a frame with a lower one, even if the reach value is the same. So stack and reach are not the best values for determining a frame's geometry for sizing.

A better method would be to locate the vertical point for measuring reach at some arbitrary stack value, rather than measuring it at the top of the head tube.