Log discrepancy - visitors

[Friend]

One of my webhosts provides a stats engine. I have been comparing my stats programs output with that of the webhosts. The page views and hits are about the same, but the webhost says my visitors are a total of 379 and my stats program says I have had 967 total visitors over the same time period.

I checked to make sure that the time period is the same (4/1 -4/8). Do the different programs use different calculations for visitors? I can understand that but this is a big discrepancy. Does anyone know what could be going on?

thanks

[dcrombie]

The webserver can't actually tell the difference between a 'repeat visit' and a visit with a long break between hits. I think the standard is to count a new visit if there is no activity for +/-4 hours.

There's also the problem of 'dynamic' IP addresses where the same visitor or spider can show a different IP address for every request (eg. Googlebot). Depending on how smart your analysis program is you can get very different numbers.

[Friend]

thanks dcrombie. I figured that there was some calculation going on with a visitor based on time between requests. Im still kind of clueless as to why the reports are so different. I assume that my software provides better analysis than the simpler included stats on this site. I may have to contact the software company and see if they have any comments.

[cgrantski]

I can think of a few possible reasons: different filtering of junk visitors, different intelligence at the DNS resolution end (maybe their software lumps all AOL IPs together and yours does not), different treatment of cookies (does your site give out cookies?), using IP/UA field combination to sessionize instead of just IP, different timeouts (WebTrends Live used 15 minutes I think; Coremetrics uses 12 hours or so, most others use 30 minutes), single hits from other sites to images being counted as visits, single hits from other sites to PDFs being counted as visits. It's always possible the hosting service has more accurate results - being in the business, they may have good software and knowledgeable setup.

[Friend]

Cgrantski..sounds like you know your stuff. I emailed support at Weblog Expert and they told me that 1. I can change the visitor timeout length (30 mins default) and 2. I can setup a filter to include only visitors who access pages.

I tried both. Increasing the timeout didn't really matter. Filtering out visitors who didn't access pages brought the visitor down to about the same level. So you're right on. I had 350 visitors who accessed pages and files, and 600 visitors who did not access any pages but downloaded a file.

I had no idea. Is this common? Or is it weird that I have so many people accessing files without visiting the pages? I do have 20-30 PDF's on the site. I guess users could be accessing those directly. I'll have to keep studying the logs.

Thanks for your insight.

[cgrantski]

That's an alarm flag for me (all the pageless visits). There are legitimate explanations (if Google etc points directly to a pdf for example), and non-legitimate ones (another site is pointing at your pdfs without giving credit). It could also be a cool picture on your site that somebody's attaching to their blog signature. It's worth trying to isolate them and looking at the referrer.

[Receptional]

More reasons for logfile inaccuracies are her [webmasterworld.com...] but the non-page view visitors passed me by on the first thoughts on that thread.

On that front, it is quite common for people to have a logo or image on the site that is called up from an advertisement (e.g. espotting type ad, or banner exchange software) on third party sites. This would mean that each time the logo/banner appeared on the third party site, another visitor would be visiting your webspace.

Dixon.

[cgrantski]

Receptional, that's a great post on that other thread!