Web Traffic Opacity

"So how many people read your blog?" It's a question I get asked, but it's actually very hard to answer. There's no standard counting method and different traffic-monitoring services give different answers. Of course in some sense it's always been the case that nobody really knows how many people read The Washington Post or watch The Sopranos, but the internet has created an odd combination that provides the illusion of server-derived precision (by one standard, at least, precisely 669 people clicked on the individual page of my "Ghost of Grover Cleveland" post on Friday, generating 844 distinct page views) with the reality that there's simply no widely accepted industry-standard counting method.

The upshot, as Louise Story writes is to substantially retard the development of online advertising and throw the whole future of media into doubt.

Comments

How do feed readers get counted? I know I read the story but did so from Google Reader and not on the Atlantic site itself.

Posted by Huntly | October 22, 2007 10:57 AM

And then there are those of us who read by RSS, and only leave hits when we come to comment. Google Reader might serve a post to 100,000 readers from just one hit in the log.

Posted by neil | October 22, 2007 10:58 AM

I've always felt that, logically, there's no reason why we can't have precise and accurate server-based counting, but that it's something that has to be deliberately built into the infrastructure of the web -- in the servers, in the browsers and in the transfer protocol. It's not something that you can try to tack on in an ad hoc way, and I suspect that during the early 90's development of HTTP, nobody cared that much.

Posted by Aaron S. Veenstra | October 22, 2007 11:04 AM

(by one standard, at least, precisely 669 people clicked on the individual page of my "Ghost of Grover Cleveland" post on Friday, generating 844 distinct page views)

But, of course, this is a good measure of how many people are interested in the comments to the post, not in the post itself, since people who are only interested in the post and not the comments can just read it on the front page.

Um, unless they came to the page through an outside link...

Posted by Al | October 22, 2007 11:09 AM

I've always felt that, logically, there's no reason why we can't have precise and accurate server-based counting, but that it's something that has to be deliberately built into the infrastructure of the web -- in the servers, in the browsers and in the transfer protocol. It's not something that you can try to tack on in an ad hoc way, and I suspect that during the early 90's development of HTTP, nobody cared that much.

As someone who worked on the early development of the web, it turns out to be a much more difficult problem that one would think. Take proxy servers, for example. Ten clients downstream of a proxy server may request a page view, but the originating server only gets the one request from the proxy server.

The obvious solution would be for the proxy servers to report their downstream statistics to the upstream servers via some protocol, but the problems are nigh insurmountable. How do you know whether you can trust what some random proxy server told you its downstream customers requested? If you decide to only trust some subset of proxy servers, how do you authenticate those your trust?

This is just a small for-instance. There are similar problems of trust at whichever layer you choose to poke at. Fact is the numbers are always going to be nothing more than a rough guesstimate, and that should be okay.

Posted by dob | October 22, 2007 11:55 AM

i'd just like to point out that i myself am a comment-generating community of several distinct human beings. that means many eyeballs, every time i click.

and then there's the fact that i always web-browse in a room full of friends who read over my shoulder.

our shoulder.

our shoulders.

anyhow, for each click, we're talking, like, dozens of separate consumers, just waiting to be influenced.

Posted by kid bitzer | October 22, 2007 12:02 PM

How do feed readers get counted? I know I read the story but did so from Google Reader and not on the Atlantic site itself.

Yes - where do I find an answer to this?

As I read through a news reader, does the original blog get credit for a page view when I read it through the reader? And, if so, how does it count it?

For instance, does the Yglesias blog get credit for a page view for each blog post I read through a reader, or just one? (Since this blog is frequently updated, I might get 4 or 5 stories at a time through my newsreader, so that if it counts by updates, it might capture 4 or 5 posts in one update)...

Posted by McKingford | October 22, 2007 12:13 PM

the amusing thing, of course, is that advertisers pay big bucks for tv ads without having the slightest indication whether anyone is watching them....

Posted by howard | October 22, 2007 12:47 PM

Yeah, I was going to say the same thing Howard did. It doesn't seem to have retarded the growth of TV advertising.

Posted by mq | October 22, 2007 12:59 PM

I
READ
YOU.

Every day.

Considering my HOT interest, one hit of mine should count for 1 000 hits in the COLD statistics. At least!

Your blog is one of my three favourites.
(Glenn Greenwald and Juan Cole are the other two I visit every day.)

Posted by Leo Brux | October 22, 2007 1:09 PM

mq, my favorite along those lines is what advertisers will pony up for a super bowl ad, despite the rather abundant evidence that most americans in front of the super bowl on their tv screen are eating, drinking, being merry, and losing track of the ads altogether....

Posted by howard | October 22, 2007 3:33 PM

My computer has a little camera in it. It is always watching me. It is watching me right now. It knows what I am doing. Don't you guys have those little cameras too?

Posted by cw | October 22, 2007 5:04 PM