Estimating Bookmarking frequency from log files

Using favicon.ico, Unique visitors & Entry pages

[stevenha]

Hello everyone,
I was looking through my access log stats, and I'm puzzled about how to interpret "favicon.ico" accesses. So I thought I'd ask if you'all can advise me ( on this minor issue ).

I know that a favicon.ico log entry is recorded whenever surfers (using IE) set a bookmark. I have favicons in my root directory and all my major subdirectories.

Here's what puzzles me: My logs show that most users "Entry pages" to my site are to internal pages in subdirectories ( >90%). Only 2.2% of my visitors enter at my home page.

But, 88% of my favicon.ico accesses are from my root directory. I presume this means that most users who want to set a bookmark, are navigating to my homepage and setting a bookmark to there. Is that the right way to interpret this?

I've also figured that 4% of unique visitors set a bookmark. I derived this figure by dividing the number of favicon.ico access by the total number of unique visitors, during the past 8 months. I imagine this is an slight underestimate, since non-IE browsers are not requesting favicons.

So, what do you think of these stats? Is a4% "Bookmarking rate" a fairly low figure?

[bird]

Two things:

Favicon.ico requests are only a reliably indicator of bookmarking when they come from IE. Mozilla and Konqueror (the browser of the KDE project) always request it when they visit a site where they don't have an icon stored yet (for displaying it in the address bar). Not sure about Operas behaviour in this point.

Some browsers (possibly depending on the version) only request favicon.ico from the root directory of the server. This is why you get disproportionate numbers for this one file relative to those in the subdirectories. Those browsers/versions that look for favicon.ico in subdirectories first will also fall back to the one in the root if they don't find it elsewhere.

[martin]

You'll probably get requests from directories if you put this in the head of your pages:

<link rel="shortcut icon" href="/favicon.ico" />

Opera doesn't use favicons, as far as I know everything derived from Gecko uses them and requests them every time they don't have it cached.

[stevenha]

Thank-you Bird, That's useful info.
So, if Mozilla (particularly) and Konqueror are always requesting it, then I basically shouldn't even rely on the favicon.ico stats as a bookmark indicator, because it can also just be a simple visit indicator.

Oh, I apologize for posting this in the Google forum. Feel free to move it to Tracking and Logging.

[Mikkel Svendsen]

Yes, you can rely on the stats like any other "internet stats" :) - Nothing is "absolute" here - Internet tracking is all about trends, not absolute numbers.

If you want your bookmark numbers to be more precise you can analyse what percentage of users are using browsers that support favicon.ico and use that in your calculations as an index number. Since it is most likely that the majority of users are using IE, and it is fair to estimate that the other users bookmark as much as IE users do you can just add this to to total numbers found.

So if 90% use IE on your site (and other favicon.ico supporting browsers) and you have 90 favicon hits the "real" number is probably closer to 100. As browser use change you can ajust for this factor and thereby get a more precise estimate - a better trend-meter

This is the same technique that is often used for unique visitor calculations too. In Denmark we usually have less than 5% users that do not accept cookies so we just add that to the total number assuming that the last 5% probably act the same way as the other 95%

From a statistical point of view this is a valid method :)