Heh, the OpenNMS project was recently on Slashdot and it took out our five year old server. I moved the website to a new, shiny server and since we use Mediawiki we added a Squid front end to improve the speed.

However, this totally pooches our "page view" counter on Slashdot since the image that Slashdot uses to count such things is cached.

I'd rather our community have a faster wiki than to have a higher ranking on Sourceforge, so I guess our project sucks now. (grin)

Downloads are like crack, addictive and ultimatly not very satisfyingly. The real metric is how well you convert those community members, downloaders and lurkers into paying customers. We've recently begun our "early adopters" program for our Enomalism Virtual Server management dashboard (LGPL). Our initial sales have been very strong including some major fortune 500 companies. We've just about recouped all our initial development costs, which in my opinion is the clearest quantitative measurement that our open source product is succeeding.

Thanks for raising this in InfoWorld, Matt.

Tarus brings up another point...download numbers are one thing, but Sourceforge rankings are entirely different. Most projects don't use all the functionality of Sourceforge.net, but the project rankings get scored as if you do. If you reset something like Tarus did, or don't use their bug tracker or mirror your subversion checkins, then your project, while performing relatively well in downloads is stuck way down the dogpile. Course, the over-eager project manager who actually has time on his hands for trivial games like this, could spend the time manually dumping in the occasional orphaned, out of process items, and suddenly their project is in the double digits.

Yet another way to put lipstick on a pig.

Hyperic has discussed this with Sourceforge before about this, and I know they are very actively trying to figure out a way to defraud this process in a fair manner. Until then though, the games will go on...

I agree with you, Matt. Customers are king. Usage traction balanced with revenues reflecting customer value is how all open source projects should be measured.

Traffic (e.g. pageviews) and downloads do not translate directly to revenue, but there are ways on coverting one into the other. Returning visitors and dependent (or loyal) users of your S/W give you influence. Influence, unlike traffic, can be assigned monetary value. You can buy influence... well, unless it's the Linux Foundation. ;-)

I think the Fedora Project (Red Hat's community linux) has come up with a pretty good solution for this problem. Don't count downloads, count how many people connect to the update server. That's a much better indication of how many people are actually using it.

Ahh. The joy of Analytics.

What is a "download"? Grabbing a tarball from the home web site? apt-get'ing from a Debian mirror? Madriva rawhide? Dag's RPM respository? SunFreeware Solaris PKG?
What about the various "unknown" mirrors - it's Free software freely distributed isn't it?

Is there a 1 to 1 relationship between downloads and users? In my case with my principal project, one user was planning on providing it to multiple hundreds to thousands of end clients - who in turn could be 1 to many actual users.
1 download == Many.

Downloads is a measure. It's not a Key Performance Indicator (KPI) or it shouldn't be if it is. What is "your" KPI will vary from project to project.

The danger with any metric is when folk misinterpret what it means or tells you. They grant it more ... credit than what it actually represents.
1 download == 1 user? Rubbish. 1 download == 1 download from server ABC that wasn't cached by any intervening doodahs; No guarantee they even use the software.

Maybe they have a tarball fetish.... ;-)

FWIW, Eric T Peterson (http://blog.webanalyticsdemystified.com/weblog) , one of the leading lights in the Web Analytics industry has been filling out the concept of measuring "engagement". Which sounds exactly like what Javier is wanting. Sadly the tool he uses to do this is expensive by even the standards of the high end of the Web Analytics industry. :-)

But the concepts are worth watching, and maybe one day we will have an agreed method and tool for the FOSS world. One day.

Steve