(ref.doc)ark 140395

Newsgroups: comp.object (Note:
 Koenig)
Subject: Re: Patterns. Was: Do OO People Really Understand...
Date: Thu, 16 Mar 1995 16:03:48 GMT

In article <[email protected]> [email protected] (Robert Martin) writes:

> >There seem to be a lot of managers out there whose philsophy is
> >``If the project looks like it's going to be on time, that means
> >I'm not scheduling things aggressively enough.''  Until that
> >attitude becomes rare, nothing is likely to help.

> I have a theory as to why that attitude has arisen.  Tom DeMarco
> paraphrases Deming in his book "Controlling Software Projects" when
> he says:  "You can't manage what you can't control, and you can't
> control what you don't measure." (or the equivalent).
> Now, I don't know about the managers you are speaking of, but if they
> are like most, they have no real program for collecting measurements
> of the software process.  And that means, of course, that they have no
> real data from which to base control critera.  Which, in turn, means
> that they cannot determine when management intervention is needed.  

I do not think that follows.  The trouble is this: suppose you have lots
of data.  How do you know if those data accurately describe reality?
That is, how do you know that you have measured everything that matters?

If you have missed something important, you will wind up managing
in a way that makes the numbers look good but will not accomplish
anything useful.  I've seen that kind of thing a lot.

> In other words, they can't see into the process, they don't know what
> is really going on, so they can't manage it.

Yes.  The fallacy, though, is in assuming that if you can measure
some aspects of a thing, you therefore understand it.

> It sounds like your managers have found one ad hoc and perhaps
> unconscious measurement upon which they have set control limits and
> thereby use as a management tool.  The measurement is "lateness".  The
> control limit: "lateness < 0", the management decision: "Add workload
> or shorten schedule until lateness > 0".

It certainly does -- and it appears to be a universal phenomenon.

> We should be able to do better than this.

Agreed.

> Most managers, finding themselves without any way to really manage or
> control a project, fall back upon their only other option: "Dictate
> the schedule and then cross your fingers."

Yes indeed.  The funny part is that on a number of occasions I have
been able to estimate the completion date of someone else's project
better than they could -- usually by multiplying their estimate by
a fudge factor that I derive from their behavior by educated guesswork.
Something like this: `These people say they'll be done in six months;
they're behaving like people who are a factor of 2 too optimistic,
so I think it will take a year.'

This doesn't apply just to software, by the way. [...]