Google by forrests


        Traditional Advertising
•   Television
•   Print media
•   Flyers (junk mail)
•   Yellow pages
•   Coupons
•   Internet ads
       Problems with traditional

• Hit and miss
• Difficult to measure success (how do
  you count)
• Difficult to correlate ads to purchases
    ADWORKS: Pay-per-Click
• Companies bid for search keywords
• The companies are only charged when a
  customer clicks on a company’s ad
• The cost per click can be anywhere from a
  cent to a few dollars and this is determined
  by having advertisers bid for keywords
• Display ranks based on those that will
  generate the highest revenue to Google
  based on the interaction between bid price
  and the likelihood of purchase.
     Advantages of Pay-per-Click
•   Pay based on performance
•   Easy to measure
•   Direct correlation between ads & purchases
•   It flattens the playing field between large and small companies
•   It puts the companies and ad providers on the same side of the
    business equation
•   High conversion rate of ~15% for PPC compared with < 1% with
•   It is cost effective – customer acquisition cost are:
•          Search          $8.50
•          Yellow Pages $20
•          Online ads      $50
•          E-mails         $60
•          Direct mail     $70
      Classroom Exercise:

• Assume that the average winning bid on
  Google is 50 cents/click
• Assume 2 billion searches per month on
• Assume an average conversion rate of 15%
• Then number of clicks is 2B x .15 = 300
• Google revenue = 300M x .5 = $150
   Improving Search Algorithm
• Knowing what the user is looking for
• Searching efficiently and intelligently
  the search space
Knowing what user is looking for
• Predicted suggestions – helping the
  user formulate search terms
• Compare to previous searches
• Knowing the user better
  – data in the user e-mail box,
  – user preferences from the user
  BACKRUB – efficient web page
       search algorithm
• BACKRUB ranks pages based on how well
  connected they are. This is determined by
  the number of links into the webpage and the
  ranking of these links (recursive function).
  The idea came from academia where by
  faculty research is ranked based on the
  number of citations of their work – higher
  number of citations indicated more read –
  more relevant – more impact the work is.
      BACKRUB methodology
•   examine how many WebPages
    address the page
•   calculate the ranking of each one of
    those pages
•   use these ranking to weigh each
•   combine to come up with a single
Problems for this Ad Bus. Model
• Potential for abuse by rival companies (have a virtual robot
  click continuously on an ad thereby raising the cost of doing
  business to a competitor
• Potential abuse by advertisers through the use of SEO’s
  (Search Engine Optimization) to optimize the page placement
  on the result page of Google by creating lots & lots of fake links
  to your website
• Someone coming with a better search algorithm
• Someone coming up with a better advertising business model
  (Pay-for-Action .. problem is it is dependent on the website of
  the payer as to whether they are able to complete a sale or not)
• Used to be the equalizer between big & small .. power shifted
  from small to big
• Need for companies to continually modify their webpages as
  Google modifies their search algorithms
          Google’s response
• Constantly monitor click behavior to identify
  faked ones
• Pay advertisers who have been victimized by
• Modify algorithm every few months to trick
• Constantly look for ways to improve search
  algorithm & test new ways to:
  – Use this ad model elsewhere
  – Improve on it
  – Experiment with others like it

To top