Printable View
Right now I am getting a 3-4x difference between Google Analytics and my host's own statistics. Is that not out of the ordinary? Do "page visits" and "sessions" mean different things to different stat packages?
Which do you guess is more accurate? I use Media Temple as my host, they use Urchin and they are a popular, big hosting service for a lot of corporate and designer websites so I doubt they would do anything bogus.
For one thing, Analytics will not count Google crawlers as visits.
I've never used Urchin, but Google Analytics is one of the best out there. I used to use it and then dumped it when I started hosting with Lunarpages... I need it back, badly.Quote:
Originally Posted by Frank Petronio
Which way does it go? Google analytics does not count image or other file downloads/views unless you embed special code for each image - a real pain. If your other software did count them, then it should give much higher numbers for photo sites.
You're right , but if they download the Web page, they've downloaded the images on the Web page, and isn't the goal to see if they find and view the Web page and how long they stay? What Google doesn't do very well is track Web script driven or flash Web pages since many pages are embedded inside the main Web page, but that can be done with the host's statistics.Quote:
Originally Posted by Ed Richards
I use Google analytics because my host's stats aren't readily accessible. I don't look into the details, only a few keys, mostly which Web pages are being viewed recently, how long they're on the page and if they leave or go to another one.
I suspect that the Google stats are more realistic as far as actual visitors versus page views/downloads/etc. since I have a pretty straight up site that is easy for Google to understand.
Which, umm, takes a lot of wind out of sails.... I thought I was getting a lot more traffic.
I've found Google Analytics isn't very good at all.
Try downloading the free versions of Xlogan and Deep Log Analyzer, I use the first to collect the data from my server once a week and the second to analyse it.
This combination gives very accurate data for free, and it's extremely useful for showing where your visitors come from, which word strings in Search Engines. It also shows you which serach engines are crawling your site. It also shows me that a huge number of visitors are trying to access a Guestbook I removed more than 5 years ago.
Google Analytics doesn't use the actual log files from your site and so gives very false reports, it's is a useful tool though.
Ian
Ian,
Why do you say Google gives false reports? Within the constraint that it only records pages with the google code in them, is there any evidence that it miss reports those pages?
Google Analytics is good as a general overview of site traffic from one day/week/month to the next.
It's NOT reliable as an absolute measure of visitors. For that, you need server logs.
Any client that does not support JavaScript, has JS turned off, or can't load the Google Analytics JS file fast enough, will not be counted by Google Analytics. When one of the sites I monitor got blogged, our server registered 70,000 pageviews, while Google showed about 7,000.
Obviously if you're charging for ads, a factor of 10 is significant. :)
YMMV.
Ed, even with the Google code added the Analytics is definitely under reading the actual number of visitors to those pages.
The raw data log which is generated for my site show things are very different, thta data is extremely accurate, as Ben says above the under reporting is by a factor of 10 or more, that is rather significant :D
Ian
How long is your average visit? I figure anything that is under a few seconds is worthless (maybe not to count, but worthless as a reader/customer/viewer) -- do you agree?
Even short visits are an indication that something is driving folks to your site. While you do not want all short visits, you would expect a bunch of them since searches are not very specific.
Yet but getting 10,000 from some Russian porn almalgamation (sp?) site is worthless to me since they don't do anything for me. I rather have 100 potential clients than 30,000 wankers like that one day.... sigh, boobs drive traffic really well.
Ah, but if you sell ads those wankers become valuable.
lol I'll pass, like I don't want those "Joyful Nude" ads like some places are running. Nothing against it, just not for me.
I doubt those Russian porn sites are looking for joyful nudes. I was thinking more about ads for custom leather and rubber clothing.:-)
Actually the vast majority of the visitors trying to spam my website are from US IP addresses, that's due to the freedoms allowed under the US Constitution which preventsspammers being prosecuted.
Once those are weeded out (around 80%) then quite a number of visitors spend between 5 - 20 minutes,and of those around 40% are repeat visitors (2 or 3 visits).
The most important data is the referring sites. This site & APUG show up regularly as referrals from my Profile etc, also from quite a number of sites who link to me, many that I never knew about. It's also invaluable knowing what pages people acess the site through.
This is all data Google Analytics is poor on.
Ian
I'm no SEO specialist but I have played around with a few analytic monitors. My experience with Google Analytics is similar to others in that it tends to consistently have the lowest values for a given day. I, like Frank, am hosted by Media Temple(I've been very happy with them) and they use urchin for analytic purposes. These urchin values are exponentially higher than google analytics numbers. I also use a download called statpress for my blog which seems to be in between Google and urchin in terms of the number of visitors, session, hits, etc. The best system I've come up with is to monitor all three over time and assess trends and not get too worked up about results on any one analytic evaluator.
Like most of you, I haven't experimented with the Russian porn method of search engine enhancement. Not a lot of qualified Russian models in Central Oregon!
Mike's Print Site
Mike's Stock Site
Another point to keep in mind is the definition of "visit" or "visitor" is quite different from one statistics package to the next.
I've been trying an Adify advertising account for the past 3-4 days and so far I've made $0.41 so I doubt I'll keep it up at $0.12 per day. Right now it is just morbid curiosity....
Google Analaytics does not "re-count" visitors who return to the site in the same 30 minute window. It also does not count visits from spiders and web crawlers. So in my opinion, it is a much better measure of "unique visits" (of the human kind) than server logs. So while the numbers may appear much lower in GA than in your server logs, they are much more useful from an SEO and usability standpoint.
However, if you are looking to track things like downloads of a particular file or image, then server logs are a must.
Even filtering out the Spiders & Web crawlers shows that Google Analytics grossly under reports visitors to a site.
It doesn't give an accurate reflection of unique visitors at all because of the way it works and many visitors coming via search engines go straight into a site bypassing the home page.
If you want accurate information that can only be obtained from analysis of the server logs.
Ian
IanG - You bring up a good point. You have to install the GA tracking code on every page of your site, not just the home page, in order to get accurate numbers for each page.
By the time you've done that you've expended far more time and energy than is required to do the job properly by just analysing your sites statistics, which will give you much more information anyway and be 100% accurate.Quote:
Originally Posted by Florida-Photo
I should add that analysis of server logs give you the information on unique visitors and whether they revisit your website, good software also filters out the visits by search engine robots etc, and gives you far better information about what search strings etc are used when people find your site via search engines.
Ian
Fact: It is impossible to get accurate statistics. At best you can expect around 70% accuracy and you should only use them as a guide.
There are hundreds if not thousands of robots, crawlers and spiders which don't identify themselves by including bot or crawl or spider in their user agent. That means they are assumed to be normal visitors unless someone has visually checked the raw log files to look at the access patterns and identified them as a bot and set their IP number in a bot list. But they change their IP numbers so even that isn't reliable and requires constant human monitoring to keep IP lists upto date.
Then AOL and others use proxy servers which gives different users the same IP number and if two users are accessing your site at the same time there is no way to identify that there is one or more users at that time.
Then if people have your site in the browser cache it doesn't goto the web to get the page again. And on and on. All web stats are inaccurate.
The best you can do is to look at unique IP numbers who have visited and then entry pages and then search phrases and how often each search phrase was used.
Awstats does all this but has one fundamental flaw in that it counts search phrase use multiple times for a single visit if its in the raw log file multiple times so even that is misleading.
Google bought out Urchin and analytics is based on Urchin. My guess, and it is only a guess, is that Google are far more likely to keep an IP list of bots upto date than any ISP or web host because they are in the business of collecting and selling marketing data. And they want information to be accurate as feasible. So whilst they may report less visits it may well be because they are closer to true user visits. You and I just don't know the facts so lets not make wild assumptions about them being inaccurate just because they are smaller numbers. Fact is web hosts couldn't give a stuff if the numbers are correct. They are only interested in server performance and bandwidth usage.
slightly OT but with the advent of ajax it occurs to me that no search engine is going index page content which is fetched via javascript http requests since most of them do not process javascript. So google analytics will probably not count those ajax file requests but it can count IP numbers which visit your site even if some them are bad bot which don't identify themselves.
The moral of the story being that if you want all of your site visible to search engines, don't use ajax scripts in them and your site will get fuller indexing and better stats reports.
But they still will won't be very accurate.
Also learn to interpret the data. For example, if the sum of all search phrase accesses plus the number of other referring sites is close to the number of unique IP's or Visits, then your stats are probably quite accurate. If they are wildly different, then you likely have a lot of bad bots, which don't identify themselves, visiting your site. But it could be AOL which sometimes uses dozens of IP numbers for a single user visit.
Most of the above does not matter much in my opinion. I found Google Analytics under counts a bit, but if you use it for a long period of time, it gives you a good measurement over time. Basically, you can see the trends of increase/decrease traffic from month to month or week to week or whatever period you need to analyze. If your visits drops 20% then you need to see what is going on? What are you doing differently? What pages were visited before that are not being visited now? You are really looking for trends and page visits or even goal completions. For that, Google does an excellent job for the right price, $0.00. If you use Analytics in combination with a good web cart, your page count will be as good as it gets.
Personal Hosting stats works off of server logs while Google Analytics is third-party JavaScript (and only counts visitors that have JavaScript enabled by their web browsers). While nearly all web browsers have JavaScript capability, the difference between those who don't and those who do are search engine spiders and other bots which crawl your site.
This is only hard if you're hand-authoring every page of your site, in which case you've already "expended far more time and energy than is required to do the job properly."Quote:
Originally Posted by IanG
This is why it's vital to have a site setup that allows for layout templating. That is, all the stuff that looks the same on each page (headers, footers, site-wide navigation, etc) is managed in one place. Adding a site-wide (or even a section-wide) change as discussed here is trivial in a properly managed site.
I'm using this tool to analyze my stats:
http://www.weblogexpert.com/
It's important, that it shows good refferers stats, which is not good in google analytics.
There are two versions. Free and paid. Paid is a bit better, but free works fine.
FWIW, I had to abandon Urchin, which I had used for years on my dedicated server, during a recent server move / upgrade. It has not really been developed since Google took it over and based GA on it, and it was no longer compatible with the newest version of my chosen server OS. If I had shared hosting which offered it, I might consider emailing their tech support to ask about long term support for it.
I really liked it before that. It always had different stats from other analytics packages I've ran but it was much more detailed, it showed about 10% lower than my other various stats packages. Urchin works best with the java code bits added into your pages, but provides truly excellent stats without it which is why I had used it.
Also, sometimes internet connectivity problems between a user's browser and google (normally the user's ISP) will significantly slow loading on pages with the google tracking code. This happens to me at least a few times a week, though if its a problem for a visitor they're probably use to it.
I'd have to second your motion for using something other than Analytics. As for offline log reading software, I'm not familiar with Deep Log ANalyzer but I have used "Surf Stats" for the same function.Quote:
Originally Posted by IanG
If you want a quick counter/stats program that provides some decent detail, and is free, you can checkout statcounter.com...they offer a 'free' invisible counter you can add to your page. It's details also include referals, etc...it only works up to xx,xxx hits but they have upgrade paid versions if you like it.