Do you have a “Fat head”? (Long tail)

As a reply to my post about how to spot a drooping tail and thus create an opportunity to increase revenue, I had a great question from Greg, which I simply had to answer as part of a separate post. The question goes like this:


I don’t think we have a drooping tail (organic keywords dataset). It seems like we have the opposite! Is there any meaning from such a chart?

This is a fantastic question that I actually cannot give a direct answer to. There is not much reliable literature available in regards to the proposed “Fat Head” – where on the other hand, there is a very well documented set of theories about the “Fat Tail” – which I would like to go into detail about later (from a “Making more money” point of view).

That said, we (as in me on Greg’s behalf) could assume that there is an “artificially” increased demand for those 4 or 5 outliners in his dataset that represents the “Fat Head” due to specific circumstances (aggressive SEO optimization on those words in particular, etc.)

On the contrary, the essential question is whether the form of the function in the middle part of the head, in the first place, can give any reliable estimation about the behaviour of the head itself. This reliability question cannot be answered by assessing this individual case; it requires the careful evaluation of various diverse cases.

Finally we should remember that there are some fundamental limitations using Zipf’s law as the “truth” for all given business models. You even see some uses of the long tail theory deliberately delete head and tail outliners.

So to conclude: Looking at the above data, I think we have a near perfect distribution and that we do definitely not have a drooping tail and the “Fat Head” is most likely “just” an effect of artificially increased demand for 4 or 5 keywords.

Cheers and thank you very much for a great question like this!

N.B. I am very open to other interpretations of the indicated Fat head, so you are more than welcome to email or comment critically on this. :-)

  • Jeremy Reynolds


    I may be taking this out of context or just be naive, but couldn’t we be looking at 4 to 5 main keywords and the rest of the tail represents derivates of these main keywords? In other words, not 4 or 5 over-SEO optimized words, but 4 or 5 main words and the rest variations of these words.

    My understanding of Zipf’s law is the diminishing odds of the exact same word repeating itself, not a similar/derivative word. Based on that I would think this is possible.

    Interesting question – thanks for posting it. This is the kind of stuff that represents the professional services that you don’t get by just buying an analytics package and turning it over to an intern.

  • Dennis R. Mortensen

    Hi Jeremy

    thank you very much for the input on this – actually – quite exciting subject. However; it is 01:19 and it is time to enter never never land/. Let me get back to this one tomorrow. :-)


  • Dennis R. Mortensen

    Hi Jeremy

    Great input and thank you very much for your “go” at the question – which I think have no right or wrong answer as of yet. However; I think we (you and I) are fair to conclude that our good friend Mr. Zipf did not have the “long tail of Keywords” in mind when doing his studies – even though it was word distribution he worked on. We also agree that he tested (remembering that this is an experimental law) for the exact same word.

    However, I do not think that that derivatives of the top keywords (the head), would create a fat head by itself. Unless of course you try to add particulars from the Search Engine Algorithms (which we could test by doing a set of filters for e.g. Google, Yahoo and MSN).

    See this is why I simply love Web Analytics – the opportunity to think! :-)

    >>Interesting question – thanks for posting it. This is the kind of stuff that represents the professional services that you don’t get by just buying an analytics package and turning it over to an intern.

    You are absolutely right! – Analysis is done by people! – no tool (and probably not the intern, at least not day one) can create magic.

    Again. Thank you very much for your input Jeremy!


  • Pingback: VisualRevenue | The 5 most used Web Analytics reports - usage study