How do we measure our users and thus persuade people it’s worth adding events?
Of course we have Google Analytics installed and we know we have had over 3,000 pageviews from over 400 unique visitors, but because we have so many feeds that is an incomplete picture.
Our Apache log shows 20,000 hits. But if several users add a feed to the same web based reader, then the reader only hits our server once.
500 hits are on our Atom feed. When Google Reader fetches our Atom feed, the user agent string shows you the number of subscribers which is very neat.
But 800 hits are on our iCal feed. Google don’t include the subscriber count in it’s user agent string like it does for Atom feeds.
When we moved from MediaWiki to our own software, the URL for the feed changed. Both URL’s still work and both are being hit by Google, so we can tell we have some users there. Also, there are filtered iCal feeds available (per group and per location) and we can see some of them are being hit.
It was suggested we give each of our users their own feed URL by adding a random number to the URL. Then we’d get hit once for each user but such a scheme would mean extra traffic for us and the readers software. Also, every time a Google bot saw our webpage it would think we had a different feed that it had to index!
But there’s a problem with all of these; they don’t count active users. A user could add a feed to a reader then never look at it again, and the reader will carry on fetching the feed.
What to do? We could make all the URL’s in our feeds go through some redirection service and count clicks. But that won’t count people who just look at the data without clicking and it could be annoying to users if none of the URL’s in the feed state where they actually go.
So how can we count users on our feeds? For now we can just ignore this problem and carry on adding features and building the site, but at some point we need to demonstrate that it’s worth taking the time to add events on our site and the most effective way of doing that is with good statistics. Any ideas?