Question re Accuracy of Google Analytics Stats - How Web Statistic Tools Work

Discussion in 'Internet Marketing' started by Fergal, Oct 19, 2010.

  1. Fergal

    Fergal
    uix_expand uix_collapse
    Premium Member
    Premium Member

    Joined:
    Nov 18, 2007
    Messages:
    10,575
    Likes Received:
    1,165
    I'm seeing very different results in Awstats (in cPanel) and Google Analytics, for the number of visitors a site is getting. Awstats is showing that I'm receiving over three times as many visitors, as what are being shown in Google Analytics. I've checked the html of the site and the Analytics tracking code is shown on every page. My feeling, based on how active the site is, is that the Awstats are the correct ones.

    Do you know of any reason as to why I might be getting inaccurate results in my Google Analytics reports?

    Thanks!
     
  2. seanstevens

    seanstevens
    uix_expand uix_collapse
    Member

    Joined:
    Jul 27, 2009
    Messages:
    638
    Likes Received:
    2
    Funny you should post this. Just noticed that on my own site GA is showing a number but elsewhere I am showing a number that is almost double.
     
  3. GekiDan

    GekiDan
    uix_expand uix_collapse
    Active Member

    Joined:
    Jun 11, 2009
    Messages:
    2,594
    Likes Received:
    218
    I kinda have the same question.
    In my site, I used to site statistics counters, Histats and Google Analytics. And from the two, Histats counter more visits than GA. But in my experience, there's only few difference, maybe a few hundreds or lesser, so I think it's no big deal.

    That's also the reason why I rarely check my GA account.
     
  4. PaulPinnacle

    PaulPinnacle
    uix_expand uix_collapse
    Member

    Joined:
    Aug 24, 2010
    Messages:
    275
    Likes Received:
    59
    It's hard to comment without seeing the site in question, but there are a number of reasons this can happen.

    At the highest level, they are using two completely different methods. GA uses JS on the client side to gather stats, while most log files use information gathered on the server side. This is an important distinction that can lead to large discrepancies in the figures.

    Any client side traffic monitor (be it GA or any of the alternatives) shouldn't be viewed as a 'total number' of visitors, but more as a 'survey of visitors' where valuable information can be obtained (e.g. should you ensure your site is IE6 compatible or can you disregard this due to low IE6 usage among your users - this can be assumed from the start depending on your target market, but nice to have figures to back up your decisions).

    It's also worth remembering that GA et al don't track spiders and bots. Log files, however, record every time a file is requested regardless of who requests it (as an example, to assist with some moderation tasks I'm currently running "xenu link sleuth" on BAF. To date, the analysis has requested +250,000 individual URL's from BAF, many of which would appear on the logs and not on the GA figures) [I'll toss you a copy of the results once completed Fergal ;)].

    The numbers can improve by ensuring that the correct version of the code is used (the most up to date asynchronous version), correct placement of the code (in the head section, just before </head>, to ensure the code fires early), the general page speed is optimised as far as possible, etc. etc., but these won't address the more 'technical' issues. If a site tends to get a lot of bots, visitors with JS disabled, visitors blocking cookies, users using proxies, etc., the figures will always suffer.
     
    #4 PaulPinnacle, Oct 19, 2010
    Last edited: Oct 19, 2010
  5. Fergal

    Fergal
    uix_expand uix_collapse
    Premium Member
    Premium Member

    Joined:
    Nov 18, 2007
    Messages:
    10,575
    Likes Received:
    1,165
    Paul, that's really helpful, thanks a lot for that. I look forward to seeing those results for BAF.

    That's a new one on me, you learn something new everyday. Paul do you know if StatCounter is also client side?
     
  6. PaulPinnacle

    PaulPinnacle
    uix_expand uix_collapse
    Member

    Joined:
    Aug 24, 2010
    Messages:
    275
    Likes Received:
    59
    You'll feel differently when I come back to you with a list of broken links and the likes and suggestions on fixing some of the issues :D (e.g. removing old signature links that now give 404's, etc.)

    I think I should probably expand on the earlier comments, as I've made it look like this is the only 'difference'.

    Even stats packages that do run server side can have varying results due to how they parse the data (e.g. AWStats vs. Webalizer), so regardless of how the traffic is tracked there will be 'quirks' and they are all a 'best estimate' rather than 'sure facts'. Using a server side log analysis system you can get an artificially inflated view of 'traffic', while using a client side browser analysis system you'll get a slightly (or significantly) underestimated view of traffic... so both have pro's and con's.

    Add into the mix dynamic ip's, spoofed ip addresses and proxies (cause havoc with server side stats - unique visitors is a fanciful notion) and all the anti-spyware, AV, firewalls, browser settings for cookies/JS/images, etc. (cause havoc for client side stats) people use, and just logging any visitors reasonably accurately is a major accomplishment.

    http://statcounter.com/help/knowledge_base/Installation.html

    They provide a far more detailed explanation of how and why the stats will differ than I managed, but they obviously 'upsell' the benefits of their technique so a slight caveat on the information there.

    Happy to admit I'm not very knowledgeable on this product and am open to correction here, but as far as I'm aware it works via the loading of a 3rd party image. So while GA requires JS to be enabled, StatCounter will log visits as long as images aren't blocked (so in theory has the potential to record higher stats).

    Both of these tools (and all similar) are at the mercy of the individual users settings, but server side stats have similar issues (all be it pushing for an over estimation rather than an under estimation) so there's no 'perfect' solution (sadly).
     
  7. Fergal

    Fergal
    uix_expand uix_collapse
    Premium Member
    Premium Member

    Joined:
    Nov 18, 2007
    Messages:
    10,575
    Likes Received:
    1,165
    Thanks a lot Paul, that's great. There is excellent information in this thread - featuring it now.

    I look forward, will be glad of the opportunity to fix them :)
     
  8. Bliss Elliosn

    Bliss Elliosn
    uix_expand uix_collapse
    Member

    Joined:
    Apr 28, 2011
    Messages:
    154
    Likes Received:
    28
    Got it.

    'total number' of visitors Vs. 'survey of visitors', that makes me confused that what is the technical difference between the two actions that a visitor visiting the site and the visitor surveying the site?
     
  9. Grumpy

    Grumpy
    uix_expand uix_collapse
    Banned

    Joined:
    Apr 7, 2011
    Messages:
    214
    Likes Received:
    62
    From what I've read, what Paul means is that numbers will vary because of many different reasons. By "survey", it's likely that he meant about the users not using JS-enabled browsers, bots & spiders, and other things that would hinder the JavaScript run by Google Analytics, causing it to fail to count a visitor.
     
    • Like Like x 2
  10. PaulPinnacle

    PaulPinnacle
    uix_expand uix_collapse
    Member

    Joined:
    Aug 24, 2010
    Messages:
    275
    Likes Received:
    59
    Grumpy pretty much answered the query, but just to expand on it a little...

    The numbers given by any form of web analytics won't be perfect. If you use server side tracking or client side tracking there are inherent flaws in the data which will overestimate or underestimate the numbers respectively. This can't be avoided, so the important factor here is just to be aware of these issues and treat the figures accordingly.

    While the figures won't be perfect, they can identify very meaningful and actionable trends on your site. Some of my comments might appear negative towards analytics data, but that isn't the case at all. I'm a HUGE fan of GA, I just want users to be aware of what the numbers actually represent. The use of survey above is basically just a euphemism for "it's not tracking each and every visitor to perfect accuracy, no tool will, but it is giving meaningful trends for the majority of your traffic which can be used to make informed and data driven optimisation decisions". Using GA to say "I had 28 new visitors this week, compared to 21 last week" wouldn't be entirely accurate due to the flaws in the figures, but saying "I had a 33% increase in new visitors" would be a meaningful trend that you could identify with a much higher level of confidence.
     
    • Like Like x 3
  11. Bliss Elliosn

    Bliss Elliosn
    uix_expand uix_collapse
    Member

    Joined:
    Apr 28, 2011
    Messages:
    154
    Likes Received:
    28
    Many many thanks to all of you, got the point.

    Here I came across an article on the same subject. It might be helpful for you and the ones searching for the same query.

    Thanks once again to all for the support and the value information.
     
    #11 Bliss Elliosn, Jun 29, 2011
    Last edited by a moderator: Jun 29, 2011

Share This Page