AWStats versus LiveStats

Which one to believe?

[dickbaker]

I'm using both AWStats and LiveStats to track my visitors.

AWStats showed roughly 150,000 visitors for last month. LiveStats showed over 250,000.

That's a big disparity. Which one should I believe?

Thanks for any replies.

[le_gber]

Could live stats be counting SE robots/spiders as visitors? I know that AWstats is quite good at sorting them out - in the latest version you have a difference between viewed and not viewd traffic where the non viewed traffics is robots etc?

are they both set to count the include the returning visitors as 1 in visits and more in returning or is live stats counting them twice in visit if they come back?

hope this helps

[dickbaker]

Thanks for the reply.

They both break out the SE bots separately, and they both show both unique visitors as well as repeat visitors.

I'm really confused.

Dick

[dickbaker]

Thanks for the reply.

They both break out the SE bots separately, and they both show both unique visitors as well as repeat visitors.

I'm really confused.

Dick

[cgrantski]

They can be working from different lists of bots and spiders, and have different definitions of what a visitor is etc. Are both using the cookie correctly, assuming you have one? If no cookie, one of them can be using a better alternative way of identifying visitors than the other (i.e. concatenated IP and UA fields). Or one of them can be resolving IPs to proxies then matching on the owner of the proxies rather than going by the IPs. They can have different visit timeout limits. There are dozens of ways they can differ behind the scenes.

This was discussed pretty thoroughly some time ago, although that particular WebmasterWorld item, which had a number of lengthy replies, doesn't seem to be indexed any more.

Somebody who can interpret reports really well might be able to pinpoint the reasons for the discrepancies quickly, because the behind the scenes differences usually are evident in the reports to somebody who knows what they are looking at.

[cgrantski]

And if they both break out SE's and bots separately, that doesn't necessarily mean that the main report that you are looking out has the SE's and bots removed from those particular statistics.

Look at two other statistics: the number of pages viewed and the number of visits (sorry, I did not notice that you are talking about "visitors" and not "visits" in your original post).This will give you an idea of whether the issue is at the filter level, the visitor identification level, or the sessionizing process. If only the visitor numbers are different, then they identifying visitors differently. If the page views are wildly different like you're seeing with visits, then their filters are very different. If the visit numbers are different, it's probably the sessionizing method, which could be tangled up with the visitor identification method.

[rfontaine]

As I have stated before, the internet is a stateless protocol. That said, you CANNOT track stats to the level most stat programs lead you to believe. Every stat program, AWSTATS - Webalizer - WebTrends - etc has come up with a method they think approximates reality. Fact is, try 10 different stats programs and you are going to get 10 different results. Thats just the way it is.

[cgrantski]

Very true! None are exact, all are approximations of reality that follow their own internal logic exactly. But a discrepancy of over 50% has an explanation that makes sense. Somewhere. And one of those statistics is closer to reality than the other.

[dickbaker]

Well, I went back and checked both.

AWStats shows an average of 4.89 pages per visit, while LiveStats shows 3.91.

Also, AWStats shows the average visit lasting 250 seconds, while LiveStats shows the average visit lasting maybe 140 seconds.

I wish I knew which one to believe.

[stapel]

What do their respective "help" files or other documentation say about their definitions of a "session"? (I don't use either package, so I don't know.) It seems to me that defining a "session" differently (is it calls from the same IP address within the same hour? is it calls from the same IP and the same user-agent, with no more than a ten-minute break between calls? etc) would give you totally different numbers.

As mentioned, of course, all of these numbers are guesses, "educated" to varying degrees. To get "exact" figures, you'd probably have to do something like require login and use tracking cookies.

Eliz.

[cgrantski]

dickbaker, you've given visitor numbers, pages per visit numbers ... what's missing is visit numbers, and/or page view (total pages viewed by all visitors). That will help.

[stapel]

According to what I'm finding on various web sites, there are AWStats "sessions":

A visit (session) is all the activity that occurs without more than 60 minutes of inactivity. A session break after 30 minutes of inactivity is the more common definition used by the ad industry but you cannot yet change this in AWStats.

...and LiveStats "visits":

LiveStats uses Session IDs to track individual user sessions for people viewing reports or performing configuration. In short, Session IDs are used to identify a session and its visitor privileges.

The term 'session' is sometimes used instead of the more popular 'Visit'. LiveStats 6 no longer uses the term 'session'. This is done not only to conform to the industry standard, but to point out that 'visits' have an administrator-customizable time-out, which was not available in 'sessions'.

A group of transactions between an IP address and the web server. The default visit expires after 15 minutes of inactivity, but this time-out can be customized.

So it looks like AWStats and LiveStats are tracking different things for "sessions" or "visits".

Eliz.

[cgrantski]

good catch. that could account for the difference. lf one ls uslng 15 mlnutes & the other is uslng 60, i'm not surprised at that 150-250 difference. dlckbaker, if you think a pause in clicki.ng of 16 to 59 minutes means the sesslon should be counted as 2 visits, use the larger number. if you thinkit should be counted as 1 visit use the smaller number.

[cornwall]

I wish I knew which one to believe.

Basically it does not matter which you believe. What each will actually tell you is the relative changes over time.

In other words, if you use one program constantly it will show you a base line of numbers and referals, then if the referals double over time, you can see your trends

Have a look at [webmasterworld.com...] its old, but is as relavant today as then.

[dickbaker]

I just tried something to see if I could figure out which was more accurate.

I have a little "banner" on my site that visitors can click on for information. It doesn't get many clicks, but enough to keep it going.

Anyway, that redirects to a CDONTS page called cdoinfo.asp. Each email I get goes into a folder in my email program.

I checked the page views in both AWStats and LiveStats for the number of times that the cdoinfo.asp page was "viewed." AWStats showed 466 for November 1st through yesterday. LiveStats showed 577, although some of those are from today.

When I checked the number of emails in the folder for November 1st through yesterday, I came up with 477. So AWStats is closer, although LiveStats isn't off by as much on page views as it is on visits.

cgrantski, I don't know what the difference between "visits" and "visitors" is. AWStats shows me the total number of visitors, including repeat visitors. Wouldn't that be the same as "visits?"

Right now, AWStats is showing 142,000 visitors (including repeat visitors) for the month through yesterday, while LiveStats is showing 213,000 for the month, including the time of this writing today.

It's still a mystery to me.

[Adam_T]

Visitors usually refers to unique I find, ie. one visitor can make 4 or 5 visits in a day for example, with an average visits/visitor of say 2.3

[cgrantski]

I'm confused now. I just looked at AWStats' live demo and don't see anything other than "visits" and "unique visitors", and nothing about "visitors including repeat visitors." Perhaps we're referring to a really old version of AWStats?

But if in fact the AWStats "visitors" numbers that you are quoting are from the AWStats "unique visitors" column, and IF you have a lot of visitors that are making more than one visit during the month, then AWStats would be expected to show fewer "visitors" than LiveStats' "visits" numbers.

[dickbaker]

I'm not terming this properly. AWStats shows "unique visitors" as well as "number of visits."

As of yesterday, I had 101533 Unique visitors, and 147,623 visits, or 1.45 visits/visitor.

That tells me--and correct me if I'm wrong--that 45% of the visitors to my site are repeat visitors.

[cgrantski]

No, that's an assumption. You could also have had, as an extreme example, 101,532 visitors who came once and 1 visitor who came 46,091 times.

[cornwall]

It tells you at it's 1.45 visits per visitor according to the way AWStats measure it.

Another tracking system will measure it differently

[dickbaker]

cgrantski: touche'

I guess the only thing to do is use a number in between what LiveStats and AWStats report.

[stapel]

cgrantski said:

You could also have had, as an extreme example, 101,532 visitors who came once and 1 visitor who came 46,091 times.

That's one busy visitor!

;-)

Eliz.

[rfontaine]

Stats programs only go so far - depending upon who or what visits your website they add up the numbers differently. For example, a bad bot can make it look like you have had alot of visitors viewing alot of pages when in fact you don't.

The truth, plain and simple, is if you are looking for hard and fast numbers you are out of luck. Anyone who says that can get you hard and fast webstat numbers are either liars or do not know what they are talking about.

[topsites]

What you are seeing is a half-dozen of one compared to six of the other, while one program has chosen to display it like apples, the other displays it like oranges.

Statistics is the science -- and art -- of using arithmetic and probability theory to work out how likely an association between sets of measurements or between sets of observations is to have happened by chance alone.
[jansen.com.au...]

Translation:
...
It really don't mean squat, one is as right as the other is wrong, but as to which one is which, that choice is left up to you. And sorry I can not and will not help with interpretation of statistics, thou I may say it is best done loosely.

So pick one, flip a coin for all it matters, and run one and only one because in the end it can drive you crazy but a real difference it doesn't make. All I have ever gained from stats are clues as to what is going on, there is nothing further can be gained from it, it really is that simple.