Google Analytics vs Reality
Posted : Thursday December 2nd, 2010

Like many others I am using Google Analytics on many sites to track traffic, and have generally been happy with it. Recently though I have encountered an issue that is making me seriously reconsider it.

In short the issue is that there appears to be a truly massive gulf between what Google Analytics is reporting and reality.

What is happening in my case is that I have a site that is running Analytics on all it's pages. This site also has a number of "special" content pages, to get a general idea the basic site consists of about 50 "regular" pages and about 750 special ones and the traffic split between is about 40 to 60. At any rate for these special pages I am doing my own logging and tracking because views and interactions and statistics on them is important for me to have close at hand. And here is where I found the discrepency.

Now the basic analytics report for November for the site claims about 25,000 page views. But my tracking and logging for the special pages alone for November shows 44,000 views! That's a huge difference and only worse when you remember I am not even tracking all the pages on the site. Based on what both Analytics reports and what I suspect in terms of the ratio of basic to special pages views, if my recorded numbers are extrapolated to the whole site I get a November total page view count of 73,000.

When I first discovered this I was shocked, Google can't really be losing 50,000 page views can it?

Well following the first rule of good development (strange bugs are your fault) I assumed the problem was in my code and began rexamining exactly what is being logged and when. I reviewed exactly what my code is doing and the specifics of what I am logging. The problem is not at my end. I compared results for a specific special page that I log to Google's and it's clear Google is simply not seeing (or recording) all the requests that I am. In one example I saw a special page that had 18 views by my count over a 2 day period. By Google's count that same page, over the same time period had 6 views. It's a pattern that's fairly consistent across the board, in all cases Google has less page views then I do, sometimes by a small number but in many cases a large gap between them.

So what is happening? Well to cover off what it is not first

• I am filtering Analytics and not seeing all traffic. No I checked, I am not doing that.
• I am looking at unique page views in Analytics and not totals. Again no and I did check.
• I am logging more views because users are refreshing in the middle of a page load (or similar behaviour leading to "false" views). Again no. I log the time and other information about the requests and when I look at the pattern it looks like normal usage. For the example listed above I can see 17 different "users" looked at that page a total of 18 times and these requests are spread out throughout the day.

As you can see even with the adjustments which I am fairly sure are too high we are still left with Analytics tracking barely half the actual traffic. And that leaves us with the possibilities that eitherhalf the site users have JavaScript disabled or blocked or Analytics is in some way pretty broken. I don't really know what it can be. The JavaScript or blocking seems to be the most palatable alternative left but I find it difficult to think that a full 50% of the audience for this site would fit into that category. I guess there is a possibility that because of the content there are really bots swarming the site at all times and that is the difference but again the usage patterns I see don't look like that and really the idea that a site with 25,000 page views is getting swamped by another almost 50,000 views from bots in every month is a bit incredulous.

The one interesting thing I have realized though from this is that may help explain some other odd behaviour I have noticed in terms of traffic as reported by Google. I have noticed for some time that the number of active sessions for the site at any one time sometimes exceeds the total number of visitors that Google reports on the site. Even when that doesn't happen though the number of visitors by session at one time for peak hours (from 8 to 5) will represent 1/4 to 1/3 of the total Google reports which is also suspicious. I mean it's not too likely that users are spending 2, 3 or more hours on the site at any one time. But if Analytics is truly missing 1/2 or more of the site traffic then those numbers start to make more sense.

I am now doing more extensive logging and tracking on my end so that I can better identify and elminate bots as well as compare my results more closely with Google's. I am hoping by Monday to have enough data to perform a meaningful analysis and maybe get a better handle on what exactly is going on. We shall see. It really is... odd.

Tags

Analytics  Google 

Categories

Internet