home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.software-eng
- Path: sparky!uunet!sun-barr!cs.utexas.edu!wupost!csus.edu!netcom.com!mcgregor
- From: mcgregor@netcom.com (Scott Mcgregor)
- Subject: Re: Estimating Large S/W Development Efforts
- Message-ID: <1992Nov19.042450.8419@netcom.com>
- Organization: Netcom - Online Communication Services (408 241-9760 guest)
- References: <1992Nov18.173255.3827@den.mmc.com>
- Distribution: usa
- Date: Thu, 19 Nov 1992 04:24:50 GMT
- Lines: 54
-
- In article <1992Nov18.173255.3827@den.mmc.com> carnahan@possum.den.mmc.com (Rich Carnahan) writes:
- >Anyone had any *real* success estimating large (>1M LOC) s/W development
- >efforts? I'm interested in hearing about any creative (sic) techniques.
- >
-
- First of all, I would point out that *real success* in estimating
- large software projects should be measured in the same way that large
- civil engineering projects are. Such projects often are successfully
- estimated to withing 50% of the mean cost and times. But large civil
- engineering companies build in plenty of safety error to ensure that
- the projcts will still be profitable in a worst case scenario.
- Software people often fail to do the latter even though they might
- successfully estimate means and variances just as well.
-
- Second of all, the collection of projects that I have looked at
- suggests that the very best predictor of project duration is the
- previous project done predominantly by the same organization,
- management or individuals. If an organization has never done a
- project of that size before, historical evidence suggests that they
- won't finish it this time either--the project will be cancelled before
- it is completed. The primary reason this seems to be so is that
- projects are more constrained by the EXPECTATIONS of the participants,
- which in turn are molded by past experience, than they are by things
- like lines of code, functionality or complexity.
-
- The evidence seems to show that the only way to "break the cycle of
- expectations" is to bring in a new manager or team. And this is only
- successful at a greater risk of conflict between the unchanged parts
- of the organization. In other words, bringing in a new project
- manager might allow the organization to do a longer project that such
- a manager has experience with, but the organization does not. But it
- may set that manager in conflict with more senior managers who fear
- that such a longer project can never succeed. And it may set that
- project manager at odds with senior developers who work for the
- manager but who suspect that senior management will eventually pull
- the plug--and so they do things the old way as a personal safety
- valve. If the project manager successfully avoids these twin shoals
- the project is most likely to complete on a schedule like their past
- projects. Otherwise the most likely result is cancellation when the
- project exceeds the organizations historical norms.
-
- If anyone has data which illustrates different results, I would like
- to see it.
-
- --
-
- Scott L. McGregor mcgregor@netcom.com
- President tel: 408-985-1824
- Prescient Software, Inc. fax: 408-985-1936
- 3494 Yuba Avenue
- San Jose, CA 95117-2967
-
-
-
-