Fog Creek Software
Discussion Board

Productivity metrics

The counting LOC thread reminded me of something I had heard somewhere but can't remember where now, or if I had just made it up entirely...

Our account is moving to function pointing as a metric for planning and we have some systems that were deemed not to be applications so no function points.  And I remember hearing that there are no valid, repeatable productivity/size metrics.  Is this anywhere near a true statement?  Not that it matters, questioning the plan makes you a non-team player around here.

Tuesday, January 06, 2004

There are, in general, no metrics that can adequately quantify anything about a person.  The only reason why Meyers-Brigs can quantify you into one of 16 groups is that the group descriptions are about as specific as a horiscope.  This is one of the central fallicies of HR-psychology, in my opinion.  Figuring out the bottom 10% of your company and firing them is about as easy as figuring out which 50% of your advertising budget you are wasting and not spending it.

Flamebait Sr.
Tuesday, January 06, 2004

That's some nice hot flamebait there indeed.


Function points are very useful if done right. Like everything that is a number, incompetant power-mad management will misuse it. But used competantly, it is a powerful tool that wields the mystical, some thought mythical power of Accurate Estimation.

Dennis Atkins
Tuesday, January 06, 2004

Accurate estimation is easy.  Precise estimation is where things get tricky.

the lackey
Tuesday, January 06, 2004

Err, surely not. Precise estimation is easy but accurate estimation is nigh on impossible, as is foo() will take 2 hours and 15 minutes to write (very precise), but potentially wildly inaccurate (as in -15minutes / +two weeks).

I once tried to explain how error (accuracy) estimates worked, but gave up when the audience (guess who) started to argue that they should be symmetric ...

To be fair, perhaps they should be, but this would require the powers to accept that mistakes, interruptions and general panic are a normal part of day-to-day programming.

David Roper
Wednesday, January 07, 2004

If I say, "foo() will take between five minutes and five years," that is accurate, but not precise.  Probably estimates that are both accurate and precise is where the difficulty lies.

surely not not.
Wednesday, January 07, 2004

Accuracy of estimation drops off precipitously with distance from this moment.  That's one of the reasons short iterations are so useful.  Never put much faith in an estimate of more than 4 weeks.  (And stick to weekly goals or shorter if you know what's good for you.)

Wednesday, January 07, 2004

*  Recent Topics

*  Fog Creek Home