Cost of Change

Clark Ching had a brief post relating the requirements churn rate (25%) to the average cost of change (50x) yielding a 13.5x increase in the time required to complete all requirement in a project assuming the original estimate didn't take changes into account (and how many do).

In another post Clark examines the disparity between Boehm's research and Kent Beck's assertion that the cost of change curve can be much lower.

Mary Poppendieck at one time wrote about there being various cost of change curves in a project where some are expensive (architectural) and some are not (GUI field positioning).

I think the Mary's assertion is easily provable by merely looking at your own bug/feature database. Do all changes have the same cost associated with them? Are all changes delivered individually or are they usually batched?

The cost of change is hardly ever uniform across all change types, so why do we talk about them that way?