Collecting Visitor Stats

Where to start?

[mikegram]

I've been authoring for a while now, but I've never bothered much with collecting info like browser type and version, counting visitors, and time of last visit.

The primary site I work with is running on Apache 1.3. Does Apache have built in methods for collecting this data or do I need cookies(which I'd like to avoid) or a cgi script to collect this info? Where do I start?

[SmallTime]

Hi. Apache has built in methods of collecting information, logs of requests (and errors, etc), but making sense of it usually means running a stats analyser program. The log may need to be enabled by your host, most will provide you with raw logs - hopefully of the refer or combined variety. (there are a couple types) From there you can either download them and run an analyser on your own computer, or put it up as a cgi program. They may offer to just furnish you with reports (analog is a common program for that, free), but I like to see the raw log, and set my own parameters for evaluating it. There is a lot of info in past posts here on the subject, spend a little time with site search. I would start by getting hold of the logs, then try out or buy an analyser program and experiment with it on your desktop, then see what configuration would be convienent over time.

[chiyo]

Ive also noticed there is quite a few GPL or free or low cost scripts (particularly in PHP which seems to be most no-brainer installations as long as you hav PHP) that you can download on the net from such places as hotscripts. Never tried them but may be worth a look.

I try to avoid anything that slows down a page at all, so we rely on raw logs, analysed locally by such things as Analog and/or (believe it or not!) in a spreadsheet if we want to do real unique or power analysis type stuff!

[Marcia]

OK chiyo, you asked for it! How do you go about the analysis in a spreadsheet?

[chiyo]

Oops!

OK..

1. open the log file as a space-delimited text file in say Excell. The columns are your variables such as time accessed, keywords, page accessed etc, and the rows your hits. (Of course,, i hear you say!)

2. then you can do anything to the limits of your program and in our case our knowledge of the program.. eg: delete all graphic hits, select only certain refereers or keywords, or pages, or search engines, sort by whatever you like using whatever levels of sorting you like, sort by IP and time to see straight away a users path through the site (my favourite - i dont know any program that does this satisfactorily), write formulaes and create a template to automatically run these analysis to a report page for later reports!

We used to do this all the time as it was far more flexible, and far faster, than any log analysis programs 3 years ago. (Heh, we also found a useful role for our accountant as he had precious little revenue to add up.)

Then we found Analog, (made the acoountant/spreadsheet guy redundant - (still no revenue to add up)) and as long as you know Analog it can "almost" do anything that a spreadsheet can do! And more than other later log analysers, many of which we have tested.

Guess you knew that already, and were looking for something smart and useful from me? Heh, I thought you knew me!