Firstly, let me state the limitations of this work. I took all the software update hits from my server log and eliminated those without system information. This left about 35,000 hits. The major caveat is that I have no way of uniquing these hits. I am definitely double-counting individuals. However, FlickrExport only hits the server when it runs so I am at least counting active uses of the plugin.
The main thing I was trying to discover is whether I should still care about Tiger. FlickrExport is nearly five years old and has a lot of Panther-era code inside it. It's code that works, but it's not always using the latest APIs and language features.
The results were not terribly surprising, it must be said, but it's good to have actual data.
Apple has been shipping Intel for three and a half years now. I no longer have access to a working PowerPC Mac, not even by borrowing one from someone.
The transition really is over, isn't it? The next thing I was interested in was the split of hardware architecture by OS version. Were these Tiger users mainly using Tiger because they haven't upgraded their machine?
I think that's a reasonable conclusion to come to. Almost nobody has upgraded their PowerPC machine to Leopard. Remember that the size of these pies is not equal: the left pie is 6% of all hits, the right is 94%. The number of hits from an Intel/Tiger machine is about the same as the number of hits from a PowerPC/Leopard machine.
Models and Form Factors
More for curiosity than to actually inform what I'm going to do, I took a look at the split across the various model groups in the Apple lineup:
The poor old MacBook Air doesn't make much of a show, but the MacBook Pro is the choice of photographers, I guess. No hits from the Mac Tablet, so I guess that confirms(!!!111one) that the new Tablet runs iPhone OS. I think.
Mac vs. Mac
Last comparison for now: form factor. It kind of falls out from the last analysis, but here's the overall split between laptop and desktop:
In terms of validity, I think these numbers are at least representative. They might be slightly exaggerating the popularity of Leopard, but I don't think they do so in a way that would change the conclusions that you draw.
Compared to the Omni Group's update stats, my data shows greater adoption of Leopard. Omni's stats average 78% across all applications, or 83% if you exclude the historically anomalous OmniOutliner.
The obvious conclusion is that if you're limiting yourself or your applications to support Tiger or PowerPC, then you should stop at once. Or, at the very least, at your next major revision.
The other, less obvious, conclusion is about hardware capabilities. Just remember that when Apple stands up at WWDC and prefixes the latest and greatest with "well, we've got 8 cores and 16GB of RAM here so...", just remember that the majority of your users likely have about a quarter of each.