what does "cache" mean in your logs?

[fom2001uk]

What on earth is this entry in my logs?

cache:71dagr3rg_qj:www.mydomain.com/

It's listed as search phrase in the referrals section of my webstats report.

Any ideas what this is?

[Robert Thivierge]

When your stats package sees:

[google.com...]

as a referer, it *knows* keyword is what the user typed in for their search.

When it sees:

[google.com...]

it *thinks* it sees a keyword, but in fact, it just mean this user browsed a cached (archived) copy of your page. Images on the page will show this is the referer. If the user clicks on a link from this "cache", those will also show this as a referer.

Your stats package isn't smart enough to parse this properly. It gets bothersome when the "cache" is mixed in with an actual search expression (which then is jumbled up). But it's a trivial issue, which most will choose to ignore.

[fom2001uk]

Another failing of WT, then.

Thanks for the reply. That gives me something better to tell a client than "just a glitch in the software" etc.

[Robert Thivierge]

This is a small part of a larger problem, that the referer string is far from simple in many cases. Stats programs can often miss:
1) Non-google.com domains (google.tld if tld is rare). A package may let you list all the domains for a SE, but that list could be extremely long and changing.
2) some Image searches
3) ip urls (like 1.2.3.4 not example.com)
4) searches within searches. So it shows the broad search, but doesn't show what it's combined with. So "red widgets" show as "widgets", giving you the idea you're doing good on a broad word.
5) Encoded keywords (like AOL which is powered by Google)
6) ToolBar "I'm Feeling Lucky" search, which doesn't provide any keyword data
7) Various combinations of search and cache. For instance, keyword data can appear as a referer to images in a page. A page with 10 images can appear in your web log ten times, each with the keywords used to find the page. You either don't see this at all, or it gets counted as 10 hits with the keyword. It should count just once.

Some of these are tough, and some are impossible. I decided to parse keywords myself, because I don't know of a free software tool that does everything. All tracking software has to accept it's gonna miss a lot of stuff.

I'm curious if any else has more items for my list, especially things which are less noticable, but might be worth handling as special cases.