Re: SLOC-wars

["John R. Strohm" <strohm@airmail.net>]

"Kent Paul Dolan" <xanthian@well.com> wrote in message
news:0b1a7c596d3d0851f6e35f3e81df82bd.48257@mygate.mailgate.org...
> "John R. Strohm" <strohm@airmail.net> wrote:
>
> > "Kent Paul Dolan" <xanthian@well.com> wrote in message
>
> >>  I'd shudder to think
> >> what the accuracy of foresight guesses at SLOC from similar projects
> >> might accomplish in setting milestone dates and such.
>
> > Your shudders would be misplaced.  Such estimates are actually pretty
good,
> > IF you go to the necessary level of detail.  Discard your preconceptions
and
> > read Barry Boehm's "Software Engineering Economics".
>
> Well, the rest of the story:
>
> As a contractor at Motorola (IRIDIUM), I watched my Motorola employee
> office mate and our common manager very carefully work out an estimate
> based on a project of similar complexity done about a year before I
> arrived, using the SLOC there as a major basis for the estimate.
>
> The result?  The two level higher-up managers read the estimate workup,
> said, "Oh, you've done something like that before!?!" and cut the
> furnished estimate in half.
>
> Not even the best estimating tools and techniques will satisfy a manager
> with a preconceived deadline.

You're right about that.  I've seen a few examples.

The key to using COCOMO is explaining how the estimating methodology works,
and pointing out how the methodology TAKES INTO ACCOUNT the fact that the
team has done something similar in the past.

This is the beauty of the COCOMO methodology, once management has bought in
to it.  The only thing they can argue with is the SLOC estimate, and THAT
you have to defend.  The effort adjustment factors are all set in stone: the
methodology takes certain qualitative criteria into account, and maps them
into quantitative adjustments.  Once management agrees with you on the SLOC
estimate, and agrees with you on the qualitative, they're stuck with the
number that falls out when you turn the crank.

The other key is to calibrate the methodology to your own organization.
This requires you to track a lot of projects at the company, doing SLOC
counts, noting the qualitatives, and then doing your own nonlinear
regression to fit an estimator.  General Dynamics Fort Worth went through
that exercise, in detail, and came up with their own COCOMO scale factor and
exponent.

The key to getting management buy-in on COCOMO is to show them the accuracy
that it has historically achieved.  COCOMO is damned good, as estimating
goes, far better than anything else that was available when it came out.
The data is in Boehm's book.