Google Analytics - search traffic accuracy?

[arthurakay]

Does Google Analytics accurately report traffic coming from the various search engines?

I have been writing a report on a client over a one month time frame using data gathered from our log files. I was curious to see how that data stacked up against Google Analytics, and here's what I found:

- Google over-reported its own referrals, though just slightly.
Our log files showed 221 referrals. GA showed 227.

- Traffic from Yahoo was also over reported.
Our log files showed 59 referrals. GA showed 66.

- Traffic from MSN / LIVE wasn't even close to accurate... off by a longshot
Our log files showed 104 combined referrals. GA showed 43.

This raises several important questions... assuming that our log files accurately track the search engine referrals.

1) Does Google Analytics include Google.co.uk (and any other countries) in the "Google" category? If so, this could explain the slight over-reporting. (same with Yahoo...)

2) Why does GA fail to track MSN / LIVE search referrals? This is critical information to many companies, and really favors Google's business. If GA is reporting significantly less traffic from MSN and LIVE, companies might choose to not advertise on those engines - a decision based on false information.

[wige]

As far as the over estimation, I think, but have not yet tested to confirm, that Google may include other engines. For example, on my site I have recorded traffic from smaller ISPs that have a Google-powered search engine. My logs show the traffic coming from the ISP's web site, but Analytics sometimes shows the traffic coming from Google. Also, viewing a cached version of your page could cause the page to not be visible to your logs, but show up in Analytics, because Analytics has no way of knowing that the user is looking at a cached version of the page, which contains the call directly to Google.

As far as MSN/Live, I have seen something similar in my own logs, and it wasn't until I created an analyzer that showed just the requests that I realized I was getting hit by a spambot. The bot that I found identified itself using the useragent for Windows XP with IE 7, and the referrer was identical to legitimate search traffic until I looked at the keywords, which were completely irrelevant to my site. I can guarantee that my site was not on the first page of MSN for "FLORIDA." These requests also all came from the same block of IP addresses which was another tip off.

[kylegetson]

I too have noticed on some of my sites small differences.. sometimes not so small.. I am currious as to what other people have seen....

javascript being disabled could have something to do with it, since the code used for tracking is all javascript... but I dont think the number of visitors with javascript disabled would be very high.

[dburdon]

Arthurakay,

analytics is a pragmatic rather than an exact science. The whole point of analytics is to create a predictive framework. Small errors should be tolerated. The question is 1. What problem are you trying to solve? 2. How can analytics help point you in the right direction.

Compare like with like. So its period A versus period B. Campaign A versus campaign B. etc. etc.

If you compare Google analytics with your log files you're bound to get two different readings. The same would happen if you compared Google Analytics with another browser based analytics programme such as Statcounter. As long as when you're comparing the performance of variable A with variable B using the same source data you should be ok.

[business-directory]

I've recently spoke to Google staff about the discrepancy between the logs and GA. We ran several email campaigns (Australian Search Engine and Business Directory, find businesses in Australia clickfind™) and know for a fact the stats should have been double. But I could not get that point across.

We've accepted that the stats are not accurate, and use our own stats now.

I don’t know why they don’t use a tracking pixel when JavaScript is not enabled????

[Charles Davis]

not sure on the Analytics question, does not agree with what we know, but am seeking help with tracking on 2,200 web sites we own, any ideas? Thanks, btw, steer clear of Go Daddy if you make a living on the web, horrible service from our viewpoint.

[nazcreative]

I agree Charles, stay away from Godaddy, just another evil corporation to add to the mix. In any event, like most things, you get what you pay for. Google Analytics isn't perfect. But your paying nothing, so no one should complain. If you want more accurate analytics, you can move on up to Omniture. But you will pay for it. They offer a very good product though, just another case of you get what you pay for.

[alan]

Hi one point more and more people are looking who owns the domain name before signing up to anything. Do you this hiding behind a Domain Proxy is a very good idea?? I.E would you do business with any company without knowing where and who they are. ???

[IndustrialWebGuy2]

I don't completely trust google analytics. Not only do they have a vested interest in making themselves look good and squashing competition from other analytics providers, some of their stats are questionable. Here are three months of comparitive stats for one website I'm monitoring using both GA and Hitslink.

8892 - Jan-08 - Page Views Google Analytics
5696 - Jan-08 - Page Views Hitslink

7422 - Dec-07 - Page Views Google Analytics
5313 - Dec-07 - Page Views Hitslink

11025 - Nov-07 - Page Views Google Analytics
7884 - Nov-07 - Page Views Hitslink


2914 - Jan-08 - Unique Visitors Google Analytics
3070 - Jan-08 - Unique Visitors Hitslink

2738 - Dec-07 - Unique Visitors Google Analytics
2907 - Dec-07 - Unique Visitors Hitslink

3455 - Nov-07 - Unique Visitors Google Analytics
3705 - Nov-07 - Unique Visitors Hitslink

So GA seems to be over-reporting page views and under reporting unique visitors. I can't remember if I blocked my IP on GA so it wouldn't register my own visits as traffic, but I do remember that it was much convoluted to block the IP in GA than in Hitslink.

Additionally, and perhaps even more important ... GA reports some really strange garbage as keywords used to find the site; stuff that is completely irrelevant to the website being monitored. And when I dig into the keywords on GA to find their source and then go to that source to look up the keywords, my site is nowhere to be found. While I know that the SERPs for keywords are dynamic, the GA results are consistently too preposterous to trust it.

But as long as GA is free, I'll keep the experiment going, for a while at least. But at some point, I'll cut it off so that big brother is not watching.

And overall, I've compared a number of tracking & analytics applications and I still believe that Hitslink is one of the easiest to use with some of the best reporting. It really blows away GA in this respect.

[NetProwler]

Another small but significant reason for discrepancy between your log files and GA is when someone reads your page from Google cache, the HTML is loaded from the cache while the image files are still loaded from your server. Your server log file will still report extra visitors.

[doj464]

Dburdon,

You are spot on!

I like analytics becuase it really helps track your adwords by position. It was interesting data to see the conversion percentages by adwords position and keyword. (I wish I could get CTR by position as well). I always use another analytics package - like Hitslink or clicktracks to analyze the logs.

Additionally, it's not just google analytics vs. hitslink... all the log analyzer and page tracking software will have different counts. We've always seen discrepancies for example, between clicktracks and google analytics, and took it as a given. Our sites have close to 100,000 page views collectively, and we've seen a 15% - 20% difference consistently between the two.

[arthurakay]

I understand the concept that Google Analytics and log files are different... I wouldn't expect them to be exactly the same.

I still don't think that accounts for the *large* discrepency for searches from MSN/LIVE...

[dburdon]

Doj464,

thanks for the compliments.

I think if you look hard enough you can find CTR by average position on Google Adwords. It would be really useful if you could also get organic results the same way.

[deepsand]

Originally Posted by alan

Hi one point more and more people are looking who owns the domain name before signing up to anything. Do you this hiding behind a Domain Proxy is a very good idea?? I.E would you do business with any company without knowing where and who they are. ???

Welcome to WPW

As a newcomer, please be advised that, as your post has nothing to do with the subject of this discussion, you need to start a new one asking your question(s).

[links2see]

I use Google Analytics in addition to Wusage 8.0 by Boutell.com. Google Analytics may not be perfect, but as others have said at least it's free. Does anyone on here have an opinion on Wusage 8.0?

[NetLinksCorp]

We have notice this for quite sometime. Although Google does a pretty good job in filtering out many spam bots – We have consistently founds it reports to be highly un-accurate!

This is no different for their adSense bot – who’s is constantly claiming that people are cheating and has been the reason for many to have their money taken away and accounts canceled.

I would only recommend using Google Analytics in the same way we use it – as a presentation tool for new customers – I finds its data of little use for anything else.

[Janna122003]

I haven't try to compare google analytics report to log file results. But If I would compare it to other online website statistics for me google analytics is better.

[deepsand]

Originally Posted by Janna122003

I haven't try to compare google analytics report to log file results. But If I would compare it to other online website statistics for me google analytics is better.

Some hosts provide log file analytics as part of the package; for example, GoDaddy's Traffic Facts.

You will find such to be a very useful supplement to any other analytics data that excludes log file data.

[qh4dotcom]

If you want to test for search traffic accuracy, what you can do is put two codes on your site...the code from Google Analytics...and another code from another site stats service like Statcounter...and then compare the two

[ushuiah]

That's also a good suggesttion. Actually I also used both Google Analytics and Statcounter and I did not encounter any problem. It works fine on my site.