Forum Moderators: DixonJones
did use stats4all but then they decided to charge (no probs there) but only if they could invoice to a Dutch address. Go figure! then I couldn't get an export of my stats!
Plus didn't report on SE hits so I'm writing something for that: [sourceforge.net...]
Used to use WebTrends Log Analyzer, but decided to start tracking user behavior instead of server activity.
I'm also using a selfmade PHP script to log all referers, just to keep informed which sites are using our products ;)
NN
but this:
[pathalizer.bzzt.net...]
is uber neat too, and I use it. draws a pretty graph with lines showing where people go within your site, but not a complete solution so you would use it in addition to another stats program.
Analog sans Report Magic
I use Webtrends Live (and really like it - you can have multiple sites there for the $ 35 / month, but the page views can cost if you have high traffic...)
I also use Log Analyzer 7.0 as the 'old standby' because I can see the actual log files.
I tried Web Trends 8 and like it, but haven't got the heavy iron yet to run it like I want to.
I also used FunnelWeb and really liked it, but couldn't reason with the vendor in order to come up with a reasonable price.
Analog I couldn't quite figure out (yet)...
Web CEO didn't seem quite stable yet when I used it a while ago - although it looks quite promising.
I like Urchin as it is so simple and seems to be reasonably close to WebTrends.
I do a lot of looking at raw logs and wish there was something better than the tool in Web Trends 7.
Chewy
CPanel 6 also has Analog, but it doesn't work right. It seems to read the same log file in over and over, so that the first month shows thousands and thousands of hits now, when there were only several dozen.
I also like to check through my raw logs. They can get unwieldy, so I end up using apple+F a lot :)
Looking at all the stats produced from the softwares available i've come to the conclusion there are only a few pieces of information i am actually interested in
1) Search engine used
2) keywords used on search engine
3) which page they entered on
All the rest that is offered like which operating system the visitor has, the time of the last visit etc etc is all irrelavant to me personally.
I think Ecommstats has the most features for ecommerce sites and selling things on the net.
Molly
It will give you a list of search engines used and also a list of keywords used, but it won't put the two togther and say which keyword was searched on which search engine.
It would be useful to know whch search terms were working on which search engine.
Log analysis beyond the most basic level unfortunately involves a fair amount of guesswork. The danger of leaving to much guessing to the software is that it's very hard to know if the results can be trusted.
Subscribe!
The basic problem with server logs is that they are, unsurprisingly, server logs. What they do is to record the activity of the server. No more and no less.
This gives rise to a multitude of spin-off problems, such as "very large logs that cannot be parsed", "logs from multiple (eg. loadbalanced) servers being uncomparable", and certainly this one "an IP can mean more than one person (company), exactly one person (dsl, own ip), and less than one person (dsl, dyn ip)"
Most striking is the use of the term "hits". Frankly speaking, I am shocked to see the term used when discussing web site traffic :o
A client request actually "hits" the server. It's the right term to use in this exact case. But nothing whatsoever guarantees that (as a result of this hit) a real page will actually "hit" a browser (and be seen by a person).
So, what we have is this situation (extremely simplified, I admit)
So, for incoming traffic (eg. SE's) logs are a nice tool to find out "what drives traffic to my site", but as soon as the person behind the browser is on the site, navigating, logfile usefulness stops.
The point I am trying to make is simply that what goes on inside the server is something completely different from what goes on in a browser used by a person.
kgoeres put it in a very nice way above:
Used to use WebTrends Log Analyzer, but decided to start tracking user behavior instead of server activity
Because of this, I personally prefer cookie-based measurement methods, especially combined with IMG cgi's to catch JS-disabled browsers.
Being from Europe, I doubt that we have an excess of the kind of ultra-paranoid (corporate?) cookie-control i saw referred by Receptional:
The real problem that I see is that cookie based systems seem to be getting less reliable as firewalls start blocking cookies
Yet we still have user-agents with disabled cookies. And we still have user-agents with disabled JS. We also have browser-sharing in families, net-cafés, and public libraries, so that one cannot always equal a browser to a person.
Still, in my humble opinion, browsers are much closer to end users than web servers are. No pun intended.
/claus
HI, yes, server logs suck.
I am using Opentracker.net for a few weeks now.
They are browser based, and from what I understand that you are saying, they focus on the human, not the IP address. My understanding is that they track page views (which are human events?) versus hits. They create cookies for individual visitors and track them over the long term. I haven't really decided how I feel about this, I think that as long as I know its happening, it doesn't bother me (as much). I can always delete my cookies.
So they track individual human visitors, and like Dan_C says:
1) Search engine used
2) keywords used on search engine
3) which page they entered on
All of which I see on Opentracker.net
I am not really interested in the stats as much though, as the click-streams. Thats the interesting part for me, in terms of being able to check out what my visitors are doing. I can actually go in and out of people's click-streams, while they are online.
This is human behaviour!
Really appreciate any help or suggestion,
cheers
Generally speaking (service type, not specific company) one of the advantages of such services is that you really do not need to do a lot of irrelevant work by yourself - plug in the code, and read the stats off a browser, that's it.
After all, it's the stats that are the interesting ones, not the pushing and parsing of some amount of mega- or gigabytes. Let alone the technical configuration of a log-analysis tool, which .. well... (lack of words)
A few words of caution, though (and this is not specifically directed against any firm):
You should always make sure that the firm you choose can document how the stats are generated.
A service like this is typically a hosted one. As such, you cannot control what they are doing like you can with a log analysis tool that you configure yourself. They will tell you something, but not all, and that's fine, but beware of firms that tell you nada.
Also, a hosted service means that for each pageview at any site operated by any customer of theirs, their own servers are going to get hit. Not "hit" like in "server hits" more like, say, a form containing information is being sent off to them each time a page is shown at a client site.
The more rich (detailled) the stats are, the more information is constantly being sent back and forth between servers. There's some math here, but i'll do it in plain words:
It means that the larger the customer base is (either large sites, or many sites, or a combination) the larger an amount of data will they need to process at any time. If they provide very rich stats (very detailled information) the processing of data will demand more raw power than a few stats will.
That's two levels of complexity. The third arises the more frequent they update. Do they run one, or a few giant batches to crunch the data, or are they computing in "real time" - the more/frequent updates, the more power needed.
- see the catch? more customers or better product means higher costs.
One way to solve these issues is to buy a lot of machines. Another is to employ some advanced math and database handling. A third is to use sampling. This is actually a subset of option two, so the last option is to wait for smoke to come out of the servers, as customers sign up.
To wrap it up i'd say: Don't go for free options if you want rich stats**. Of course You can, but stick to such services that only provide one or two key metrics and does not update "on the fly". They tend to be more stable in the long run.
(** I know that the service You use is a paid one, this is generally speaking)
A firm providing "rich" stats needs to get cash in order to be able to serve you reliable stats, otherwise it faces option three.
Plus: Even though you're paying, and even though you might even have a small site, it's the total load of all pageviews on all sites using one specific provider that counts. The firm may choose to use sampling, that is: Process only the x'th of each request. That way, it's not the full picture that you get.
Having said that, it might be a good and reliable picture anyway. All sites are different, and sampling of large sites is generally not as obscuring to results as sampling of small sites.
(if you have only two pagewiews and omit one, you will have a sample of one page when sampling 50% - if you have 200PW you will have a sample of 100 when sampling 50%)
This might sound as if i have second thoughts now, but really i haven't. I still prefer this method over logfiles anytime.
Regarding cookies:
I haven't really decided how I feel about this,
- do make a privacy statement. Be honest about it. Tell your audience that their visits are being monitored and do not forget to tell them what the reason is.
You might also like to point out that a cookie is not a program, and that it cannot really do anything. Try showing them the content of a cookie. You could even provide them with guidelines on how to remove cookies.
Really, this is in your own best interest. If people do not know why they are being monitored, or if they find out "by accident" they get nervous. If they are told in plain text what happpens and why, they get familiar in stead.
At the end of the day, the purpose of monitoring web traffic is either:
a) to enable you to offer better services to them, or
b) to help you make money which, in turn, enables you to offer them better services.
(somewhat idealistic view, I admit)
So, really, this is not exactly a thing to keep secret. And they will discover anyway. The "smart users" tend to speak and write emails as well as the not-so-savvy.
If users get nervous about your site... well, do i need to finish this sentence? I'd prefer the familiar ones :)
/claus
...did i say "a few" words...hmmm...
NetTracker Quote 127,000 US$ ++
Webtrends Quote 165,000 US$ ++
and many others i checked were quite pricy. I think I've decided to let my techies write me a primitive one. To get me through the host migration.
Thanks
If I was able to muscle up the money here in Canada where our dollar is not worth a whole lot, then it had to be good *grin*
Happy Canada Day everyone!
itrainu
