Fog Creek Software
Discussion Board




Software Errors Cost Billions

(In a recent Reuters article...)  In other news:

Vehicles Kill Thousands, and
Water Drowns Hundreds, and
World's Children Cost Billions

Not to mention gems like, "software users contribute about half the problem," where the stat could easily be, 95%  or 10% depending on how you define "software error" and "problem."

And, "Currently, over half of all errors are not found until ... during post-sale software use."  Cripes!  When do these auditors start counting bugs?  In pre-alpha, alpha,  Beta?  Not at release, because then the stat should be, "all."  Over half?!  I've released much software, and I have roughly 20 post release bugfixes, out of hundreds and hundreds of prerelease bugs.

USDoC's NIST, indeed.  What pure tripe.  New headline: "NIST Wastes Money."

Ahhh.  Fun rant.  Have a nice weekend, all!

David Blume
Friday, June 28, 2002

I didn't include the link to the original article, because it was pretty easy to find.  But here it is, in case anyone thought I was making it up:

http://story.news.yahoo.com/news?tmpl=story&u=/nm/20020628/tc_nm/tech_software_dc_1

David Blume
Sunday, June 30, 2002

This is the NIST paper:
www.nist.gov/director/prog-ofc/report02-3.pdf

random java drone
Sunday, June 30, 2002

Software itself costs maybe trillions.  What's worse?

Bella
Monday, July 01, 2002

java drone,

Thanks for posting the location of the actual paper. Over 300 pages but 125 of them are appendices.

Anyway, there's lots of interesting stuff in there.

First I make the observation that the numbers throughout the report are largely wild speculative guesses.

And the conclusions are more random speculation.

I wasn't even sure if they were or were not saying that fixing errors was cost-effective!

But I did like this one hard number:

"In 2000, total sales of software reached approximately $180 billion.
Rapid growth has created a significant and high-paid workforce,
with 697,000 employed as software engineers and an additional
585,000 as computer programmers."

The workers were further broken down elsewhere:

"Computer programmers 585,000
Computer software engineers: applications 380,000
Computer software engineers: systems software 317,000
National total 1,282,000"

Now not all programmers work on software that is *sold* -- quite a few of them work on projects for the use of the company they work for. But if every last programmer was working on software that is sold than each one generates $180 billion / 1,282,000 = $140,406 worth of sales revenue.

Hm, that's not much so it must only be a small fraction of programmers that are working on for-sale software. Anyone know how many?

But then again that $180 billion doesn't even begin to represent the cost savings that the software purchase represents.

When my mom worked at general dynamics, the art department took a team of 20 six months to produce a 200 page manual. This involved several hundred thousand dollars worth of printing and photographic equipment as well, though that is spread out over several projects.

Using MS Word, it takes me less than a month to do a 200 page manual singlehandedly. Total cost of equipment - maybe $1500. And $200 for a copy of MS Word. So because of software, a $130,000 1965 job (in 1965 dollars) now costs a few thousand 2002 dollars. Those dollar amounts are a result of the programmers too, not just their trivial $180 billion in actual sales.

Anyway...

About that 1.2 million total programmers.
That's not enough jobs for everyone in the US to live off of. So the idea that we are becoming an information economy and we don't need to manufacture stuff anymore, I think is false.

Also, didn't I see somewhere that the total number of unemployed and underemployed programmers in the US was OVER 1 million?? Why are we even bothering to fund engineering programmers at public universities -- it's obvious we have a massive oversupply of them and that is depressing wages.

X. J. Scott
Wednesday, July 03, 2002

Oh yeah, note that the method they use is calculating the cost of fixing bugs in a waterfall model.  Bugs found later are many multiples more expensive then earlier.

This is precisely what Agile-type engineering methods are designed to counter.

random java drone
Friday, July 05, 2002

*  Recent Topics

*  Fog Creek Home