Friday, October 8, 2010

CEDO 540 - Week Four -


This past week our class delved into the world of market research and Internet data.  Particular emphasis has been on Bill Tancer who is the GM with HitWise.  It is one thing to know what sites people currently visit but Tancer takes it to the next level by gleaning data on what people are seeking on The Internet.  Bill's company pays ISPs to aggregate data on what sites Internet users visit and, of greater interest, what search strings they are using.

Hitwise's bread and butter is providing data on Internet traffic combined with reporting.  Customers can purchase any number of reports showing upstream/downstream traffic from their Web-sites compared to those of their competitors.  Hitwise helps subscribers better understand the demographics and lifestyles of Internet users.

Bill Tancer is now investigating ways to relate Internet users' search strings as a predictor of their future behavior.  One of Bill's celebrated examples is when he attempted to predict the winner for a Dancing with the Stars competition based upon the raw number of searches for each contestant.  It almost worked the first time, but what he had not factored was that one of the contestants was 'searched' for in a context unrelated to the competition.  The contestant, Stacy Keibler, was a  wrestler and men were searching for her pictures so they could admire her beauty - or something like that. 

The Stacy Keibler scenario reminds us that the science or methodology for mining the intent of search strings is in its infancy.  It still comes down to humans analyzing the data and questioning metrics used to reach conclusions.  Being human, we have a natural bias.  In his article, 'Lipstick and the Narrative Fallacy', Bill Tancer refers to the inclination to only see the data that supports a hypothesis and ignore the data that might disprove it.

Tancer cites the widely held view that purchases of cosmetics increase during tough economic times.  In 2008 he analyzed the number of searches for 'lipstick' and found that they indeed had jumped during the corresponding economic downturn.  But, upon further review, it was the complete search term 'Lipstick on a Pit Bull' that drove the number up for 'lipstick'. The data could have supported the cosmetic/economy theory, but by looking at the complete picture it was just Sarah Palin with a great quote!

5 comments:

  1. I was wondering this week if the Stacy Keibler effect also happens in the classroom. I ask students questions all day and I'm basing my opinion of what and how much they're learning on their answers. What if I'm not asking the right way or if there is a misconception between what I mean and what they mean. It just reminded me to ask deeper and higher order thinking questions rather than superficial, just the facts type questions.

    ReplyDelete
  2. I was once told by a cosmetics and jewelry sales woman who was trying to get me to also peddle her snake juice, that since we were hitting a recession, sales would go up. I did a little research and evidently it's true. http://www.populationstatistic.com/archives/2008/11/15/lipstick-economics/ I guess I should have bought into her scheme :(

    ReplyDelete
  3. Sarah, perhaps cosmetic sales go up as more people are likely to get roped-into peddling to their friends? Hmmm....interesting to correlate the number of people selling during a bad economy vs. good.

    ReplyDelete
  4. Brian,

    I liked your analysis and thought your example of Stacy Keibler brought a good point. I do not think there will ever be a "fool proof" way of using Internet data. There will always be a need for humans to find out the real story. We can see all of the humans using those skills to attack their opponents in political ads!

    ReplyDelete
  5. I think the Stacy Keibler effect in the classroom may have more to do with boys thinking about long legs. Using data is always needed to be done with caution. We can often get a bit over confident with what we think we know and it really is subject to our methods and the intervening factors.

    ReplyDelete