Without any "client identifying things" how would GA be able to chain several page hits into a session then? That is, do basic visits vs. hits split.
If you are in fact anonymizing everything about a client as you claim you do, then it won't be able to. Unless, of course, you are feeding GA some opaque client ID that you then internally map to and from actual clients that hit your server. However something tells me that you aren't doing that, or you would've mentioned it already.
(edit) I re-read your comment. You aren't apparently interested in session counts. But what's good the GA summary then if you can't tell 10 bounced visitors from one visitor with 10 hits? This makes no sense. If you want to look at just page hit numbers, there are dramatically simpler ways to do that.
In the test I've done, sending no session/user data over, I lose all sense of a "session".
But I do retain insight into what content has been viewed, how much, what is rising and falling, etc.
The question really is what info are you really reporting on? AdBlockers make us blind and tracking is horrible, but I get to have a far more complete view over the simple stuff Urchin used to be great at.
Ah, so you are passing some client IDs over the GA after all. An IP address perhaps? You know that's a leading question, right?
Incidentally, I ran similar experiment with gaug.es few years ago - pulled on their tracking API from our server side. While it worked as expected, these sort of shenanigans are good for only one thing - hiding the fact that you are using 3rd party analytics from your visitors.
On a more general note - the thing is that you either care about other people's privacy or you don't. It's not a grayscale, it's binary. And if you do, there's no place for GA in the picture.
I am not passing IP. I am not passing a client-id. I am not passing any kind of correlation identifier from which a session can be inferred or created. I am not passing user-agent information. I am not passing a cookie ID.
I am only passing a page view event. "Page /foo/bar?bash has been viewed".
But isn't the same kind of data you could extract from Apache logs? Since from what you describe is basically a log of all your requests.
GA has many utilities, mainly is to follow the user and see the funnel they go and second to monitor the marketing campaigns. If you don't need this, then Apache log + webalyzer is perfect for everyone.
I persist with GA, because every now and then I work with partners who would like to verify the activity on my websites (and yes my user agreements and privacy policy allow this) and have a means to compare this with historical data or data from other sites.
Those partners frustrate me, in that they won't trust me to provide stats generated from server logs, but they all default trust GA.
This technique allows me to use GA, produce the view of the content they need, export the PDF, and share that... and they trust it.
GA is the de facto store of trusted data when it comes to web site activity. For my sites that is tracking content page views.
This whole conversation started with you saying why abandon GA when you can use it without compromising clients' privacy. An exchange that followed shows that one can't actually derive not just the same function from GA that way, but virtually any function at all. Yes, you can feed data in, but the usefulness of what you can get back out is next to zero. What am I missing?
From your opening comment:
> Why not move to push GA data server-side?
Because it renders GA largely useless if clients' privacy is actually observed.
> I am only passing a page view event. "Page /foo/bar?bash has been viewed".
I would like to say, as someone extremely hostile to tracking of any kind, that if this is all you're sending to google, that sound perfectly fine from a privacy perspective. (Google gets your information, but that's between you and Google)
Thank you for choosing a method that respects the privacy of your readers.
> (edit) I re-read your comment. You aren't apparently interested in session counts. But what's good the GA summary then if you can't tell 10 bounced visitors from one visitor with 10 hits? This makes no sense. If you want to look at just page hit numbers, there are dramatically simpler ways to do that.
I do not care to track users/sessions, page views are enough for me. I am tracking content and content views... and I get this big tool that is awesome at slicing data and presenting trend information... for free.
If you are in fact anonymizing everything about a client as you claim you do, then it won't be able to. Unless, of course, you are feeding GA some opaque client ID that you then internally map to and from actual clients that hit your server. However something tells me that you aren't doing that, or you would've mentioned it already.
(edit) I re-read your comment. You aren't apparently interested in session counts. But what's good the GA summary then if you can't tell 10 bounced visitors from one visitor with 10 hits? This makes no sense. If you want to look at just page hit numbers, there are dramatically simpler ways to do that.