Do you use Google Analytics? Well, a recent study shows that you’re not alone, roughly one third of the Alexa top 500 websites traffic wise use it. That surprised me a little. Anyways, I use Google Analytics myself and noticed before that some numbers are inaccurate or downright wrong. In May I discovered how wrong, completely wrong.
The numbers of search engine referers, more specifically, are wrong. I’ve noticed that before but wasn’t panicking as no statistic tool is perfectly accurate, you always should use at least 2 of them and compare the numbers. Now this time I checked my stats over at SEO 2.0 I was unable to ignore it anymore.
SEO 2.0 is mostly about social media and blogging SEO and it relies on traffic from other sources than Google. I only get insignificant numbers of visitors from Google on my blog, mostly due to very specific “long tail” queries. The only really important keyphrase is “SEO 2.0″ itself. Now looking at my May stats I was very surprised that one of better performing keywords was “site seo vs blog seo” without quotes.
I looked deeper into that just to discover that all 60 visitors looking for this keyword is one person from Bucharest, Romania. Google Analytics counted her or him several times a day for more than 2 weeks.
Now I sometimes create returning visits just by launching my FireFox with my saved tabs but how can a person search for the same query every day several times and then click the search result each time? This one riddles me. I’ve seen that kind of behaviour with Google Analytics in the past but always thought it was due to myself or several developers from my client working on the same site.
Not this time, there is certainly no Romanian developer working on my site. Now try to multiply these numbers for one of the top 500 sites and you end up with completely worthless statistics.
I’m not here to bash Google for their great analytics solution. I love checking and comparing bounce rates etc. but it’s more than strange that a search engine gets the search referers wrong in its own analytics solution. So if they don’t want people to make up conspiracy theories about Google overstating their search referers numbers they better fix this quick.
What can you do about it? Look out for some alternatives and either use Google Analytics and at least another solution or even more. What alternatives are there?
Well, if you’re running a “top 500 website” you better rely on a solution used and recommended by leading search marketers. This is ClickTracks in most cases. Also Omniture has been recommended by many professionals. Many people outside of Germany probably don’t know it but some German marketers prefer Etracker. Just to name three of them. There are plenty of high end solutions out there you surely can afford if you run such a huge site.
For the average webmaster seeking a low cost or free solution there is no single alternative to rely on. There is Piwik, an open source alternative recently hailed by many, but it’s in alpha now and didn’t work for me, it produced SQL errors instead.
Another solution is to use Google Analytics along a specialized search analytics tool like:
Have you experienced similar issues with Google Analytics or am I the only one to discover this?
There have been a post at SEOmoz, by SEO theory blogger Micheal Martinez already in 2006 and a post in 2007 from AimClear focused on the reliability on Google Analytics with another topic, outages.
My conclusion right now is a grim one: Google Analytics is cheating you, I hope not on purpose. Unless you do not care about accurate search referers and visitor numbers you should use one of the solutions above or resort to log file analysis.
I’m afraid what you describe as “wrong” data is simply your interpretation of a normal feature in Google Analytics as indeed virtually all Web Analytics packages worth their salt.
Let me explain:
What happened in this instance was that our Romanian friend discovered your site through the specified keyword and then returned on every other instance “directly” (i.e. by typing the URL outright or through a bookmark link).
The rationale here is that if someone discovers your product through a given source of traffic, thinks about your offer for 24 hours, makes his mind up and then returns to your site by typing your URL directly and converts (as he now knows your site thanks to that wonderful keyword), then it makes perfect sense not to overwrite that initial referrer info and to attribute the conversion to it.
This only happens with subsequent “direct” visits. If the following visits has Yahoo! as a referrer, then of course the previous source will be immediately overwritten and the visit will indeed be attributed to Yahoo!
This isn’t some weird conspiracy on behalf of Google but rather what makes sense form an online marketer’s perspective.
Agree? Disagree? Doesn’t really matter as Google Analytics gives you the possibility to change this option: add _ucto=”18000″; to your page tags and the previous referrer info will only ever be kept for 5 hours (though instead of 18000 put whatever value you want in there in seconds) and all future visits by Romanian boy through his bookmark will appear as “direct”:
_uacct=”UA-xxxx-x”
_ucto=”18000″;
urchinTracker();
Google Analytics rocks dude even though it may not always be the best-documented tool in the history of mankind ;)
I reported this effect a few months ago… a visitor comes to the site from a search engine referral, and is then always counted as another search engine referral every time that person comes back in the future… even if they were direct type-in or bookmarked traffic on their return visits.
Alan, that’s very interesting stuff and would certainly explain where these numbers are coming from.
I am massively addicted to Google analytics…
Its a tricky one, but I am inclined to think that although Google cannot be more than 90% accurate… its still a good indicator.
Also its possible the Romania persons IP address was not static and so recorded as different people. Also never underestimate how obsessive people can be. I know I check Digg way more than 60 times a day sometimes…
Peace
I use HotTracker for WebAnalytics and I´m very happy. It´s FREE up to 20.000 PIs per month, realtime and gives me correct data. It´s available in german as well and I like it more than etracker.
Google Analytics uses almost the same method to count returning visitors as most other webanalytic solutions do. The Romanian user you mentioned probably has deleted his browser cookies (manually or automatically) . I didn’t experience any serious trouble yet (ok, the loss of ecommerce data some weeks ago), that other solutions would have done better.
I tried piwik as well – but it’s just not nearly as comprehensive as GA. Maybe in some months… but not yet.
You could give HotTracker a try – or wait for IndexTools (hopefully) to become free as well.
Yahoo! will make IndexTools free, they haven’t announced when yet but I’d expect this to be before the end of the year.
While all the explanations you offer sound interesting there is a real problem and it persists. The 60 = 1 is just the most blatant case.
I took a look at Hottracker but the presentation does not work with newest Flash plugin and the site has not even screenshots.
I’ll certainly will hail a free IndexTools. Also I’m testing Woopra since yesterday and it’s a very viable Google Analytics alternative.
Hey Tad,
Since last month I am also facing this problem that Google Analytics showing wrong data i checked my conversions in backend there is 12 conversion in a day and same day i checked it in GA i found 0 conversions.
In alternative of GA GetClicky is good analytics application.
Another package to consider is crazyegg.com …could be useful to monitor specific pages that receive low volume traffic. The heatmap facility is useful.
Microsoft have got their adcenter analytics in beta too.
John
I’ve experienced a slightly different problem with Google Analytics where for a new site I suddenly received a few hundred “direct” visits in one day. There’s certainly no way that these were human visitors.
Despite this I do think that all in all Google Analytics is a good indicator of your site’s traffic. There’s just perhaps the odd error creeping in every now and then.
Another great analytics is Indextools, they’ve been bough by Yahoo in the last months. It’s paid, not that expensive but it is really good.
I’ve tried a few Google Alternatives but always come back to Google Analytics.
I thought i was the one facing this wrong statistics problem :-0 I have made a list of alternatives http://d95.com/f6dc I am using woopra and get clicky !
Analytics just works so well but I have noticed it’s not the most accurate. Server logs have the best accuracy.
I never notice of this issue before. I have to check again the accuracy. I agree with you, better have alternate beside analytic.
Thanks for sharing.
Most webmasters fails ranking in Google Results
because they don’t follow all the SEO suggestions
Google Site suggestions are important on the selection of keywords.
On the search bar of the new browsers firefox, or IExplorer with google toolbar.
As you type, Google will offer suggestions keys to navigate the results.
Almost any website, can have a good ranking in google following this instructions.
Put attention to this keywords and use it on your websites,
and this will drive more search Engine traffic to your site.
More tips at
http://Tecni.Com/en/seo.htm
http://ArchiveNational.Com
I’ve tried a few Google Alternatives but always come back to Google Analytics.
I found this site precisely because I had a suspicion that something wasn’t quite right with my google stats.
I’ve just got a new e-commerce web site going (about a month) and google indicates that for the past week, there were 0 visitors each day. But, there are people ordering items on those days (taking into account time differences). If these people haven’t visited my site (according to google), then how are they managing to place their orders???
Doesn’t quite add up in my books.
I like the presentation and analytic options with GA, but I suspected there were discrepancies with the accuracy.
I recently left the company, but managed web sites totalling 200 million page views per year. I ran a full year comparison with GA, WebTrends with SDC, and WebTrends using logs.
I looked at daily traffic for the entire time. I ran specific scenarios on private sites before they were launched, where I knew the usage.
We followed all of the listed recommendations. With the money we spent in this area, I had a developer dedicated to analytics.
What I found:
1. WebTrends with SDC was the most accurate in collecting information.
2. Google Analytics captured only 53% of the amount listed in WebTrends with SDC.
3. WebTrends using logs was also inaccurate.
WebTrends SDC and Google Analytics use a javascript to call the server to record a visit. This is better than log based analytics, as many people use ISP’s that cache data (meaning a page wouldn’t be pulled from the server, and there would be no entry in the log file). Same thihg happens for people that work at businesses with proxy servers. Cached page means they use the local page, and not the server page.
Script based solutions such as GA and WebTrends with SDC get around this by making the browser call the server to denote they viewed the page. What I suspect is happening is security software or firewalls are blocking the scripts from being able to call the site (or the server is blacklisted). WebTrends script was better written, and I don’t think their site is blocked.
We decided on WebTrends with SDC as our final solution. GA does some neat analysis, but I’d suggest that if you have a real business behind it you need a better system behind it. Free is only good if it works.
That’s just my $0.02.
Visit the link I used here and you’ll get some insight into how Google Analytics tracks what it tracks (and what it doesn’t track well). If you’re can afford it and are huge with an IT staff go with Omniture. If you’re a medium-sized business you might consider WebTrends; however, Clicktracks is better than WebTrends and excellent for small businesses with limited budgets too. If the price is out of range, a reseller may be able to split the cost with you.
I have tried other programs, though now use Google, too. I presumed that they had some data redundancy. I still find it valuable to know what they see and are tracking for my sites. After all, it is what Google thinks that counts. I will look into some of the tools you have suggested. TY!
One annoying thing that happens in GA often is that when I visit my own site through a keyword search in the Google search engine, GA often indicates that I visited the site up to six or seven times that day based on that one keyword phrase, but I know that I only visited the site on that keyword phrase once. Then to convolute the data even further, GA often indicates that I visited my own site up to two or three days in a row through a Yahoo keyword referral, when in fact I absolutely know that I didn’t.
I’ve seen some other strange data in GA that I have no explanation for except for maybe dynamic IP addresses or something strange happening on the end of the ISP.
The thing that first got me interested in Google accuracy is unreported visitors. I have a statistics screen within my web-host that shows all the the visitors to my site.They showed visitors that Google wasn’t showing, there were 5 countries that Google didn’t mention.These 5 countries were sending visitors but Google never mentioned their presence whatsoever.It has me a little worried however I still appreciate Google making this tool available.
Thanks for this excellent article.
I too am very interested in this subject and have just run part 1 of a test to try to quantify the delta between GA and others that measure traffic in similar ways.
How many page-views are you really getting? Part 1 – The WordPress Stats Test
http://www.synaptici.com/2010/how-many-page-views-are-you-really-getting-part-1-the-wordpress-stats-test/
Hope this is of interest and would love your feedback.
Cheers,
Gil
It’s great that Google provides this for free, and I feel like I’ve gotten value from it in the past, but I found this article today because I Googled “what is wrong with google analytics.” Right now, it’s showing that one of my client’s sites has had 0 traffic pretty every day in the past few months except July 23, when they supposedly had 2 visitors from Brazil. Their hosting company’s stats say differently. And I know I’ve visited their site several times in this period, and I’m not in Brazil! Very unusual.
thanks for the list, i use http://hitstatcounter.com/ myself as it uses piwik and you dont need to install it on your servers.
I got 5 comments today and GA shows zero visitors.
Just found this thread and very interesting. I’ve just installed some real time tracking software on my site. One term that appears in the live stats as driving traffic to the site doesn’t appear on Google at all. The live stats software can’t be making it up! Also, yesterday, the live stats reported 189 visitors. GA reported 96.
Very worrying!
Andy
Hi, the infor here is great and Alan’s contribution too. We’re having another kind of problem.
We get a visit from someone in a remote town (say in the US) on a given day. His first connection counts as one visit but then it seems that each time he clicks on another page, it counts as an additional visit.
Result: we have fifteen visits on the same day, from a remote town, each visit has a 100% bounce rate, one page displayed only. Only one has a specific source and the remaining visits have ‘direct’ as source. I am sure these are all the same visit…. Does anybody have this same issue? Thanks for your help.
I started to add my own tracking code next to the Google Analytic script. Simply count page views, place a session cookie to count the visits, and a permanent cookie plus counting the IP’s to get a sense of unique visitors. Put the results with PHP in a MySQL database and count the results at the end of the day. I then removed the webspiders from the results. The total number of views, visits and unique visitors counted by my script is around 30% higher then Google reports. My stats seems to match the results from the server statistics as well.
I can imagine some differences in results but this is extreme.