Log Stats Accuracy

[nfzgrld]

I'm going nuts with my stats. I have three programs I look at: Webalizer and Awstats for my log files, and Web-Analyse which is a php app that keeps it's own log. I get three different results from them. In particular, the daily unique visitor numbers are all over the place. Webalizer doesn't even have a daily unique visitor stat. For yesterday Awstats gives me 54 and Web-Analyze gives me 89. Which one, if either is accurate? Who knows? What you you guys using, and how can you tell if it's really accurate?

[J.D.]

Visit analysis is based on a timeout, which is the time between the very last hit and the one before. This timeout may be different for various analyzers and usually ranges from 20 to 30 minutes. This timeout may be what makes your stats different. For example, if a person from 127.0.0.1 looked at your website at 12:00 and then at 12:21, the analyzer that has the timeout configured at 20 minutes will count the second hit as a new visit. The analyzer that has a 30-minute timeout will count the second hit as a part of the first visit.

On top of that, some analyzers use IP addresses to ID users, while others use cookies. Cookie-based approach is usually more accurate. For example, two people from behind a firewall will be counted as a single visitor if IP addresses were used for identification and as two visitors if cookies were used.

J.D.

[nfzgrld]

The two analyzers I'm using the most both use IP for identification. Also, I can update either of them to the current time by clicking a link. AWstats can be updated to within the last minute, and web-analyze can be updated to right now, or there abouts. Also, remember I'm looking at unique visits, so you're 20 minute scenario really isn't applcable.

I think what I need to do is pick one and go with that. I basically just want to track stuff from one day to the next just to see how I'm doing.

[J.D.]

Originally Posted by nfzgrld

Also, remember I'm looking at unique visits, so you're 20 minute scenario really isn't applcable.

I think you are mixing up visitors and visits. It sounds like you are talking about visitors, which is the number of unique IP addresses or cookies for each time period. It's easy to verify which analyzer is more accurate. First, extract all IP addresses from your daily log and then sort them:

sed -n -e "s/<ip-addr-regex>/\1/p" | sort -u

Then just redirect this into a file and look at how many lines you have. This is your correct visitor count. If you don't know how to use sed, PM me a couple of your log lines and I will help you with the regular expression.

J.D.

[exam]

Originally Posted by nfzgrld

Also, remember I'm looking at unique visits, so you're 20 minute scenario really isn't applcable.

JD was explaining how to define a unique visitor.