My Photo


  • Clicky Web Analytics

Your email address:

Powered by FeedBlitz

« Evaluating Internal Search Performance By Topic | Main | Wagging the Dog: More on the Great comScore Debate »


I agree with you: this is no trivial matter. I think we all reacted like Anil, and still want to believe in our hearts that, well, trends, trends, trends. But this study, if really rigourous in its methodolody and findings, is really going to bring a lot of nerve racking dicussions. And we should have them: a visitor is going to have to be a person soon. What are we going to do when database and analytics integration is on every marketer's lips? How long can we live in a world where our visitors are likely to be suffering from multiple-personality disorder,and change identities when they commute?

I think you do a fine job in putting the percentage in the context of the web site type, but still, you're right to be very worried. I know I am. I am even wondering if getting back to same-session analysis won't offer at least more solid data... but we know that's not what we want to do.

Are we to say that maybe the problem, the anonymous nature of our data, has its solution in the "soft personalization" of the good'ol days of greeting visitors by their name? I know Amazon still does it, and I don't need to login. In fact, I hate deleting my cookies exactly because I end up not being recognized by the sites I use frequently. Which means that they can at least link me to some kind of profile.

I know, this sounds quite simplistic. But, again, we will need to get closer to the person, and people will only accept it with sites they have a vested interest in. I think one day we will determine acquiring an email, or anything else that permits starting a dialog, as valuable a conversion as a sale itself. And when we start knowing WHO's coming to dinner, we will care less about the anonymous part of the trafic, because of the tremendous value we will be able to extract from visitors who are willing to say hi.

Well, you will excuse my disorganized thoughts: hard to express all that in a comment.

this is perhaps the most reasoned response I've seen to the issue.

My concern over the study revolves around the fact that it seems to be a single internet entity and a single ad network. Even if it is one of the top five, I wouldn't call it a representative sample in terms of the types of sites on the internet.

Furthermore, I don't really find the argument that 'comScore is up to something' argument. I might be inclined to believe it if the study had come from Nielsen - who has a competing site analytics product, but as far as I know comScore doesn't and isn't interested in competing in that (our) market.


I want to make it clear that views expressed on my blog are my own and do not represent ZAAZ or any other employee of ZAAZ.

I have posted a response to you post on my blog


I commend you on these two thoughtful postings.

I must admit that, when we conducted this study, we were primarily focused on the inflation in UV metrics and in campaign reach estimates provided by ad server logs. The issues you are bringing up are obviously very important and I am not surprised at the debate they are generating. We are in the process of writing up a more detailed paper on our analysis which I expect we will release next week. I hope it will answer a number of questions people have about the study specifics. In the meantime, let me clarify a couple of things relevant to this dialogue.

We chose the ad server and large portal cookies to get the broadest reach possible and provide the broadest view of cookie deletion behavior. In particular, the ad server we chose is so ubiquitous that you can hardly surf the web and not get their cookie. This is useful because it provides a good proxy for overall cookie ‘resets’ by the user, since a cookie reset will almost always translate into a new cookie for this server.

Obviously individual sites will be impacted differently. In general, the impact increases as a function of site’s frequency of visitations. At one extreme, a site that is only visited once during a one month period will not be impacted at all since it will have only one cookie per user regardless of how often the user deletes cookies. On the other hand, a site with a high degree of repeat visitors will tend to count a higher fraction of its visitors multiple times. The duplication is more acute if the site’s audience skews towards people who have a higher propensity to delete cookies.

As I think about the issues you raise, it occurs to me that it may be possible to come up with empirical adjustments to new and repeat visitors as well as campaign conversion metrics generated by Web Analytics software. We can specifically track the deletion rate of the site’s own cookie(s) and derive the appropriate correction for the various metrics.

I have always believed that audience measurement services like comScore are highly complementary to Web Analytics services. In that we tend to be focused on providing added value about competitive information and demographic profiling of the individual user, whereas Web Analytics go much deeper in analyzing the site’s own audience. This study may in fact imply that the two types of services are not just complementary but also interdependent. Each service enhances the value of the other, particularly if these correction factors are used. I would be curious whether you or your readers agree.

The comments to this entry are closed.