Weird Google Analytics

[Jason Hansman]

Hey everyone,

I was looking at my Google Analytics this week and as of Wednesday they went absolutely crazy.  My social network averages between 150 - 200 visitors a day; now I'm getting statistics that there are close to a thousand a day visiting - we peaked at 1,500 on Friday.  This seems suspect at best as my community only has a little over 5k.

What's even more interesting is that my pageviews has stayed constant (1,500 - 2,500):

And my bounce rate has skyrocketed from 25%-35% to 80%:

Has anyone else had this problem?  Or even know what's going on?

Thanks!

Jason

[Andrew Krzmarzick]

To everyone seeing unusual increase in direct traffic - this is an error by Google Analytics (not Google's fault or your fault - most likely caused by Ning adding a 2nd set of Google Analytics code to your pages). Every page with this double code is calling GA twice (once to your account, once to Ning's account). The 2nd call on each page is rewriting the cookies Google uses to record your traffic in GA, thus erasing all past history of each visit. Every pageview beyond your first visit is being counted as a completely new direct visit because the cookie containing the visit and true traffic source information is being overwritten.

You'll see pageviews on your site roughly equal visits on your site (this will be true for every page that has this problem - look for pages that have same number of visits and pageviews, those are your problem pages). You'll see pages / visit creep closer to 1.0. You'll see referral and search traffic remain roughly the same as you would expect, with your entire jump in direct traffic. You'll see bounce rate creep to 100% and Time on Page creep to 0:00 on your problem pages, because every page view is a new visit and every visit ends on that page (without two successive page views within the same visit recorded on your site, GA can't calculate time on page and gives it 0:00).

The problem:
1) Ning's GA code is set up (using Asynchronous GA code)

var _gaq = _gaq || [];
_gaq.push(['_ning_ga._setAccount', 'UA-2270114-6']);
_gaq.push(['_ning_ga._setDomainName', 'none']);
_gaq.push(['_ning_ga._setAllowLinker', true]);
_gaq.push(['_ning_ga._setCustomVar', 1,"member_type", "anonymous", 3]);
_gaq.push(['_ning_ga._setCustomVar', 2,"appId", "4126342" ,3]);
_gaq.push(['_ning_ga._setCustomVar', 3,"language", "en_US" ,3]);
_gaq.push(['_ning_ga._setCustomVar', 4,"product_type","premium" ,3]);
_gaq.push(['_ning_ga._setCustomVar', 5,"member_count","361" ,3]);
_gaq.push(['_ning_ga._trackPageview']);

2) GA information is sent to Google by the Ning code (Asynchronous)

(function() {
var ga = document.createElement('script'); ga.type = 'text/javascript'; ga.async = true;
ga.src = ('https:' == document.location.protocol ? 'https://ssl' : 'http://www') + '.google-analytics.com/ga.js';
var s = document.getElementsByTagName('script')[0]; s.parentNode.insertBefore(ga, s);
})();

3) Your GA code sends the information to Google AGAIN using the Traditional GA code (I assume?), likely rewriting the cookies because you're using the Traditional GA code in conjunction with the Asynchronous code which causes problems (not your fault, Ning's fault for adding Asynchronous GA code and not warning everybody).

<script type="text/javascript">
try {
var pageTracker = _gat._getTracker("UA-xxxxxxx-x");
pageTracker._trackPageview();
} catch(err) {}
</script>

The Solution - Use the Asynchronous GA code on your site - this should "play nice" with the code Ning is using:

<script type="text/javascript">
var _gaq = _gaq || [];
_gaq.push(
['my_ga._setAccount', 'UA-xxxxxxx-x'],
['my_ga._trackPageview']);

(function() {
var ga = document.createElement('script'); ga.type = 'text/javascript'; ga.async = true;
ga.src = ('https:' == document.location.protocol ? 'https://ssl' : 'http://www') + '.google-analytics.com/ga.js';
var s = document.getElementsByTagName('script')[0]; s.parentNode.insertBefore(ga, s);
})();
</script>

With this implementation, you'll still be making two calls to GA, but because they're both Asynchronous it should be okay. I don't think Ning will let you set up your GA info within their code, or you could just add your info before they call ga.js and not have to call it twice. Make sure you delete your old Traditional GA code when you make this change.

I haven't verified that this is the golden solution, but it's my best guess given all the evidence, and is now what I'm implementing on my site as of last night. Time will tell if it's working.

Hope this helps.

[AnitaC]

I don't understand all the language you are speaking. Can you break this down in layman's terms, i.e., what direct effect is this having on us each day, week, month. My stats were looking awesome until this happened. I have almost 7,400 members on my site, have been up for a little over 2 years and now this. So what happens to our OWN Analytics that we use - are they useless now? Can I rely on my data?

[John]

I replaced my Google Analytics script with the asynchronous GA code.
It didn't correct the problem.

Appreciate you posting your best guess Andrew.

[J.Wings]

wondering if you saw any changes... I didnt, but it was good to update to the latest code anyway.

[david kralik]

Haven't noticed a problem on our end. Page views seem to be correlated with visitors. Although, I did see a spike on Monday, June 14th and again on June 16th.

Since june 7, bounce rates have increased 30% (and remained steady) at that level.

[MAXX]

It woulda been nice if ning had said something about tracking our pages.....guess they got caught red-handed.

[MAXX]

Thanks for explaining this so thoroughly Andrew. If only ning had taken the time to do the same......

[Radialistas Brasileiros]

I got exactly the same scenario! Your stats are looking pretty much like my stats!

And the most intriguing is the direct access source for the most of the visits!

Uh, and I was thinking my site was starting to going viral, hahahahahaha!

[Radialistas Brasileiros]

But I don´t know... the number of pages served in Adsense grew up the same proportion of the Analytics stats, did this an error too? So Adsense stats showing it´s serving more ad spaces is not real?

I don´t think so. I also was thinking that with the implementation of Caffeine my pages is more better indexed and letting more visitors, but how if they are coming from direct source and 90 plus % newers!

[Dave Andrews]

For the Sony sites they are seeing the same thing. It started on June 9th.

Bounce rate is up and duration is down. We were trying to figure out if they did something on air to drive traffic to their sites and they said they didn't.

[John]

I trust Ning didn't know our stats would get wacked.

[Lora]

Does Google Caffeine have anything to do with the spike I am seeing

http://www.technewsworld.com/story/Googles-Caffeine-Perks-Up-Search...

My traffic has more then doubled according to GA... 58 % in fact