Friday, September 16, 2005

Tracking my blog's "customers"

Peter Scott's blog recently looked at problems with identification of customers, and the first response remarked on how much harder it is with internet 'click' customers as opposed to 'real' people walking about. I'm going to talk about what I know, or think I know, about my blog's customers.

A blog's customers are its readers and despite (or because of) being new, I've kept an eye on my site stats to try to get a feel for my customers. Since I left my SiteMeter public, you can see them too, just by clicking on the link to SiteMeter at the bottom of the blog or here [go on, have a look at what I can tell about you :) ].

My item on Jonathon Lewis's book got triple the traffic of any other page though I don't think he should get too excited since we are still talking double figures here. But for me trying to understand my readers, it is interesting. Most of my visitors, where I can tell, have come from orablogs or a similar blog aggregator. Although there were links about the book on AskTom's front page and Jonathan's own page, neither featured in an RSS feed, so it was probably news to most readers. Of course, I've got no idea how many readers OraBlogs have, so I am probably judging by too small a sample.

Since my previous internet activities were in forums, my blog extended from there, being mostly technical and resembling answers to questions which hadn't actually been asked. While I'll still do that, the 'Jonathon Lewis' spike gave me a feel for my 'customers' suggesting that blogging, and being syndicated by OraBlogs, is more like having a chat in the pub (or coffee-shop) with your workmates.

The Oracle WTF [] definately has that "You'll never believe what I saw today..." feel and other blogs have newsy items such as what has been released, who is going where and so forth. I'm therefore going to try to throw some more non-technical talking points out and see what happens.

Judging by my SiteMeter data, especially with my relatively small traffic figures, its feasible to track down details of readers and especially those who leave comments. I wanted a counter at least to pick up hit numbers and was a bit surprised by the amount of information it recorded, especially with the referrer details. I think I can turn the referrer details bit off and/or remove public access from my site traffic. However I don't want to get into some of the heated discussions that can go on, and I don't expect people to post anything that they wouldn't say 'aloud'. Since the information is recorded I figured I might as well let people see what is collected about them. My own 'Freedom of Information' policy if you like.

I'm open to persuasion though ironically you'd need to leave a comment here to voice your argument. Obviously you can use one of the anonymiser sites, and I think the referrer bit can be blocked by stopping Javascript and/or following the instructions here for Mozilla/Firefox.


Jeff Hunter said...

Yes, it's certainly interesting to know who your readers are. I've learned a lot about my readers; both technically and personally.

Robert Vollman said...

I also got "tripled" when I was included in OraBlogs.

I've gotten minor spikes when someone like Niall or Pete F linked to an article.

I didn't notice that much of a spike when I first mentioned Oracle WTF (which was before you, Bobablog and Jeff :P :). Except when Niall linked to it and I got the "Niall effect."

I hear the ultimate "spike" in traffic is the Tom Kyte Effect.

Jeff Hunter said...

I hear the ultimate "spike" in traffic is the Tom Kyte Effect.

I didn't have stats before Tom linked to my blog, so I don't know. I can tell you that about 1/3 of my "came from" links are from Tom's blog. I got a spike from when HJR linked to me.

Noons said...

Yes, but the blogsphere is a particular case where the referer info is important. For search engine marketing and add placing with p4p and a few other google/yahoo "niceties", the really interesting bits are the timeframes, the address/area of the users and the search terms. Plus assorted cookies for verification and such.
Quite an interesting arena: did you know there are 365 common used mis-spellings for searching "Britney Spears DVD"? Amazing...

BTW Gary, I'm in Cremorne even tough it shows up as Waverton. Where are you nowadays, after OSR? ;)

Gary Myers said...

I've picked up some hits from search engines (surprising as I don't think I'd be highly ranked) but you are right, as a blog they're not something I'm interested in.
I can understand misspelling Britney, but DVD ? DUD, VD ? Are you sure they don't mean to use those terms.

PS. Got another spike when I mentioned possible changes to AskTom. Might change my blog to a Tom Kyte/Jonathon Lewis gossip column. Get the paparazzi up to Toronto for their next meeting.

Noons said...

"I can understand misspelling Britney, but DVD ? DUD, VD ? Are you sure they don't mean to use those terms. "

Heh! Tell me about it... Didn't believe it either, until one of our sales folks showed me. Note: typos, although unintentional, are also a valid search expression and the results get clicks as well!

One of the interesting aspects of all this search engine stuff is actually figuring out what people *mean* when they type a search term, where do they end up and what they really wanted.

By the time all the permutations are figured out, it's quite common to manage a customer targeted campaign for a range of terms in excess of two hundred thousand. And that's for a medium site. These all then need to be monitored for effectiveness and efficiency, with a view to reducing the cost to the punter.

The software to manage and capture the results of all this has to go through oodles of data every day and extract some sense out of it. Lots and lots of regular expression stuff, custom coded. Mostly C, but also Perl and Python. With Oracle as the end storage mechanism for further analysis. Quite interesting.