Tools or No Tools
We are currently working on a Special Report on Application Quality and Testing which will be posted in the next week or so. I have spent a good part of my career in the development of code quality tools, and remain mystified by their general lack of use by developers. For example, I spent time on the BoundsChecker (
http://www.compuware.com/products/devpartner/visualc.htm) team (as product manager) a few years ago. BoundsChecker is probably one of the best-known and most effective of the C/C++ error detection tools. While we attempted to promote regular usage by developers, by far the predominant usage pattern was as a last resort to find and diagnose a known bug.
Many developers I've talked to use only the IDE (Visual Studio, Eclipse, or other), with no additional tools for unit testing, code coverage, error detection, or performance. With the possible exception of unit testing harnesses such as NUnit and JUnit, most do not even bother using available free or open source tools.
Software development is a profession of trade-offs, so let's look at the tradeoffs here. Let's say that the average developer with a few years of experience makes $90K (some of you may find this number out of reach, but I know plenty of developers who make more). Adding benefits, facilities, and applicable taxes, a fully loaded developer probably cost 50 percent more, or $135K.
Many of these tools claim to be able to increase productivity finding and fixing bugs, or improving performance. In many cases, the tools vendors claim up to a 50 percent improvement in productivity. That is probably exaggerated, but 20 percent is not unreasonable, especially since quality tools can identify and diagnose problems that sometimes cannot be found any other way.
Some studies have indicated that developers spend perhaps 50 percent of their time debugging. This might be a little high, but for the purposes of working with round numbers, let's use it. This means that tools can possibly be worth ten percent of a developer's loaded salary, or $13.5K.
That pays for a lot of developer quality tools, with some left over. So my question is, why do so few development teams make the investment? I had always thought that development teams were more willing to invest in additional bodies than in tools, but the cost of tools is trivial compared to the cost of an additional developer. And how do you explain the lack of use of free tools, especially in the Eclipse community? Do we think we are so good that we don't need additional quality tools?
I am not planning on going back to work developing quality tools any time in the near future, but I am curious. Any answers that you might have would be gratefully accepted.
Posted by Peter Varhol on 11/05/2006