Those all-important ratings – now for the web!

The Nielsen TV ratings system people who bring you wildly inaccurate numbers to show who’s watching what on TV have a system for rating websites. Unfortunately, like the TV system, it relies on subscribers. They need people like you to download a surfbar that tracks your movements online so they can record where you go. The problem is:

  • Surfbars were invented in 1996 by companies who paid you a few cents a minute to let them track your web movements.
  • Scripts that trick them into thinking a human is surfing naturally 24 hours a day were invented, oh, a few months later.
  • Alexa has been using a surfbar to determine web metrics for years, and even though it gained acceptance because it was the only one widely available, any webmaster who owns more than a few sites can tell you it’s way off. (BTW, Alexa is by Amazon – it’s that thing that enables them to recommend stuff they think you’ll like. Which provides millions of people with good laughs daily.)
  • I once downloaded the Alexa surfbar, surfed Hathor heavily for three months, and boosted Hathor’s ratings into the coveted top 100k. What a joke.
  • Predictions are already being made that Nielsen’s will go the same way because, as with TV, getting real data from a non-random sample of volunteers who are free to lie to you in several ways is about as informative as standing out in the open in orange and making a lot of noise in hopes of observing the natural behavior of nearby gorillas: you might see natural behavior, you might not. You’ll never know.

But is that even what they want? Granted, compiling TV ratings is a challenge for anybody, but web metrics? There’s a much better way to do it: instead of relying on unmotivated to download stuff which motivated webmasters can also download and cheat with: ask webmasters to add a tiny bit of code to their pages. That enables a system to track two different measurements and presume the truth to be somewhere in between. Still not perfect, but far better than anything Nielsen has ever come up with.

This is how Quantcast works. It surveys the audience for a hopefully random sample of data, then asks webmasters to insert a bit of code for tracking. I inserted the code for Hathor a few months ago because I wanted to see how our demographics really trend, since I’m sure 90% of the people I knew in film would be certain our audience is mostly female, mostly old angry bitter feminist dinosaurs. Now you can see for yourself, and track the numbers in the future.

Our Quantcast stats.

The age breakdown

As you can see, our audience does indeed toward the coveted 18-34 age group, but with a strong presence in the 35-49 age range, too.

And what do you suppose our gender breakdown is? That 18-34 year old group couldn’t include many males, could it? I mean, we know they don’t care, right?

The gender breakdown

Dead heat. Hmm. Well, don’t worry, I’m sure the target audience boys are only visiting us to, um, make sure we don’t hurt ourselves with all these electronics we’re using to produce a website and read it. Yeah, that’s the ticket.

Race breakdown

The race breakdown

Wait! Maybe the visiting young males aren’t white! That would explain it!

Damn. Well, they must all be gay! Gotta be! I mean, even though we have male posters who are married and unmarried ones who blither on about hot actresses in comments… um, it must be a trick! “Curses, you evil feminists – foiled again, but we’ll be back! With newer better rationalizations!”

I may be joking around in this post, but whenever a TV show or movie performs differently than expected, people rush to supply rationalizations for how their expectations really were correct, but something went wrong. Anyhow, now I’ve put my money where my mouth is: Hathor isn’t about a bunch of crazy feminists wanting something else no one cares about. It represents a small but growing audience of people including those advertisers and movie makers claim to want most in their audience, who are sick and tired of the same old same old.


  1. says

    How does it tell if I’m male or female?

    I mean, I obviously have a female-identified name, but “BetaCandy” isn’t exactly gendered, nor is Purtek.

  2. says


    Short answer: it’s an extrapolation based on the survey data.

    Long answer: the demographic data comes from their surveys. They get demographic info from a sample of people, then track where those volunteers surf. That much is the same as Nielsen’s doing.

    But then they also collect the info from the tracking code on Hathor, which does not tell them gender, age, etc. but does tells them where people are coming from, what time of day they visit, etc. They bounce that against their survey data and look for correlating trends like (and I’m just making this one up):

    –From the volunteer sample, we know people in Arkansas who log on after 8 pm tend to be over 50, white and female.
    –Site A has a majority of people logging in from Arkansas after 8pm.
    –Therefore, Site A has a majority of people who are white, over 50 and female.

    Does that make sense?

  3. Mecha says

    I’m a big fan of well done statistics… but after buying a bunch of books on women warriors and sexuality based on various recommendations, Amazon thinks I’m a lesbian, which is a little off. Do you think stuff like that might skew your stats? ^_~

    (I know it doesn’t, but I couldn’t resist. That is a neat way to try to get decent correlation statistics.)


  4. says

    Hey, Mecha – remember the Alexa thing I mentioned in the article? That’s Amazon’s puppy. That’s why their data is so far off. I shall edit the article to point that out.

    Like Nielsen, it relies on people downloading surfbars and using them without agendas. It probably also has a survey group for correlation, but gets no direct info from the websites.

  5. Mecha says

    Huhm. I wasn’t under the impression that Alexa data helped feed the main Amazon site’s ‘like this? try this!’ logic. It seems a bit too far afield to be of strong use.

    It tends to be ‘funnier’ when it ends up making assumptions like ‘you might be a lesbian because you read books about bisexuals, transsexuals, and the history of women warriors’. Or, alternately, I have a male friend who bought a pink DS off of Amazon… and was subsequently treated like a marketer might treat a 13 year old girl in recommendations.

    I think there is some interesting analysis space there, both statistically and societally. Is Amazon’s buy data accurate? What does a ‘wrong’ recommendations imply about the algorithm? The people involved? Should they be looking for more cues? What happens if they get them? (And that doesn’t even get into the game of ethics of data mining.) Fun stuff. Anyway.


  6. says

    Mecha, it’s my understanding that Alexa is not the only component – they also make extrapolations from what you buy and what other people who bought that thing also bought. But that doesn’t tell them what they need to offer that they don’t already have – for that they need a sample pool letting them know of their other interests online. (And where else you’re buying stuff.)

    Again, that’s just my understanding.

    I think their extrapolations probably ARE based on assumptions about what your interests say about you, and that’s a series of leaps marketers just don’t want to quit making.

  7. Genevieve says

    Ha. I love those recommendation things on Amazon. Apparently after the holiday season last year, my mom kept getting recommendations for books about feminism and mountain-climbing: I’m into the former, my sister’s into the latter. Who knows what comes up if you combine the two?

  8. says

    I should’ve waited until today to post this: we just took a leap of 5k in total visitors and are now ranked within the top 100,000. And we’re trending even more male now. 😀

  9. Dunvi says

    Both. First of all, that they can know who’s what gender anyway, and that the right hand number for Asian seems to be just about equal to Caucasian but the percentage is far far lower.

  10. says

    Dunvi, that number is an average index number for the whole net. What it means is that Hathor has more than the average number of Asian visitors for a site.

    As for how they know people’s gender, race, etc., see my reply to Anna above, which explains it a little better. :)

    As for the fact that they track it at all… well, I’ve said for years that demographics is the new racism. It makes a science out of assuming you can predict someone’s behavior by their race or gender, but it all boils down to the same thing. :(

  11. Dunvi says

    Ah, see now that makes much more sense. I was starting to worry I was only worth, what, a twentieth of a caucasian?

    Oh, and demographics are crap. Although it is sort of fun scanning the internet for numbers that look good and sound credible.

  12. says

    Eileen, if you have separate logins, I believe it could tell you apart by storing a different set of cookies in your browser than in his, but if the two of you actually use the same copy of the same browser, then anything working like Quantcast or the Nielsens will think you’re one hybrid user. 😀

    Another thing Quantcast can’t detect? Whether someone’s visiting a site because they love it or hate it. Like the Nielsen’s can’t tell if your TV is tuned to something because you’re avidly watching or because you got bored 5 minutes in but left if on after you went out to mow the lawn.

    Not that they’re trying to get accurate numbers.

  13. says


    Since I wrote this article, our stats have jumped. Back then we were getting around 12,000 visitors per month. As of today – less than a month later, we’re at 47,700. We’re now trending 62% male with a considerably stronger showing in the 18-34 demographic than in the 35-49 (which were previously about equal, as were the genders). And we continue to trend predominantly white.

    In simple terms, Hollywood’s target demographic – white males 18-34 – forms the majority of The Hathor Legacy’s audience.

Leave a Reply

Your email address will not be published. Required fields are marked *