Question:
Difficulties with parametric estimation techniques
A common area of inconsistency with parametric estimation techniques such as COCOMO is due to the fact that quite often task completion times for software development are not additive. For example, if coding up a new module, the time to design, program and unit test the item may be rather high (say a few days). However, additional similar modules could take less time (say a matter of hours) due to familiarity with requirements and re-use of existing code.
Do you, have any suggestions for how the effect of extensive code reuse can be factored in to parametric estimation methods?