duncan@geppetto.ctt.bellcore.com (Scott Duncan) (03/28/89)
As the topic of software cost estimation has come up (even if only in terms of
a requested reference to COCOMO), I thought an ACM (Communications) article
from last Spring (May '87, v. 30, #5, pg. 416) might be of interest to those
looking at cost estimation models. It is by Chris Kemerer (MIT Sloan School)
and is entitled "An Empirical Validation of Software Cost Estimation Models"
and looks at COCOMO, SLIM, ESTIMACS, and Function Points. It concludes that
"all of the models tested failed to sufficiently reflect the underlying factors
affecting productivity."
This is not to say that one cannot make use of such models (or products derived
from them); however, the general wisdom on this subject is that you need much
local project history to "tune" the models to your environment. Some feel that
you may as well develop your own models based on your own history data since
they are going to be the best.
Another model, SOFTCOST, developed by Robert Tausworthe at Jet Propulsion Lab
has been mentioned by some as an effective (and more "generic") model than some
of those mentioned in Kemerer's article. It does, I am told, require input of
67 variables which is enough, in some cases, to discourage project management
from using it despite its usefulness.
Speaking only for myself, of course, I am...
Scott P. Duncan (duncan@ctt.bellcore.com OR ...!bellcore!ctt!duncan)
(Bellcore, 444 Hoes Lane RRC 1H-210, Piscataway, NJ 08854)
(201-699-3910 (w) 201-463-3683 (h))