On March 16, 2010 I attended the SDForum meeting of the Business Intelligence SIG at SAP in Palo Alto. The topic was "Web Analytics 2.0" presented by Avinash Kaushik, the author of Web Analytics: An Hour a day and Web Analytics 2.0. Not only was the subject matter of this presentation interesting and informative but also the speaker, Avinash, is quite a dynamic and interesting. I was informed and sometimes even entertained by Avinashes stories and analogies. This is one of the best SDForum meetings that I've been to so far.

The the slides from the presentation can be downloaded from the SD Forum Business Intelligence SIG page. I've included a direct link below:

The following are my notes from the evening.

  • Announcements


  • April 9, 2010 - one day conference on web analytics
  • Speaker: Avinash Kaushik (@avinashkaushik)

    • TV, radio, magazine advertising are "faith based" and cannot be
    • New face of marketing is the conversation based thing like twitter


  • Co-founder of MarketMotive
  • Blog: Occam’s Razor
  • Web Analytics: An Hour a Day

    • $80k of author’s proceeds went to charity


  • Web Analytics 2.0
  • Now an Analytics Evangelist at Google
  • Rethink Analytics
  • How much more democratic is decision making now and how much different is it.
  • Not a single Business Intelligence was successful in web analytics

    • The things in BI that make them successful is not he same as web analytics


  • Web Analytics 2.0
  • Used to be able to blame everything on not having enough data but not anymore.
  • Free web analytics from Yahoo. Yahoo Web Analytics

    • It can do everything you need


  • We’ve been thinking about web analytics wrong

    • Clickstream - The What
      • Multiple Outcomes Analysis - The How Much (quantify the revenue and conversions)
        • Experimentation & Testing - The Why
          • Voice of Customer - The Why (Why do the customers do the things they do, you cannot guess so you have to ask them)
            • Competitive Intelligence - The What Else (There is a ton of data that you can get from your competitors, mostly for free)
              • Insights - Insights












  • Multiplicity

    • In BI you could build the single source of truth but you cannot do that anymore on the web
    • Use multiple tools for different tasks - you can’t use just one tool


  • Ideas Worth Sharing

    • Rule #1: Don’t Stink
      • Don’t suck because if you suck then nothing else matters
        • HITS - How idiots t--- success


      • Bounce Rate - "I came, I puked, I left"

        • It’s a very actionable metric
        • You can cross reference other metrics with bounce rate


      • Look at your own home page. Look at the top pages on your site have high bounce rates, then go fix it.
      • Landing pages matter, need to be appealing and not suck

        • SAP can;t be found on


      • Landing page for Qvinci looks a little more appealing - a page that doesn’t suck (small business reporting searches)


    • Segment or Die

      • If you report data in aggregate, then you thing all of your users are exactly the same
      • Reality is that they are all very different
        • You can’t analyze individuals so you analyze in segments


      • Look at pages per visit

        • Figure out what the people who are engaged/loyal (multiple pages) are doing


      • Can do on the fly segmentation with any of the tools
      • Problem w/a content rich website/homepage is you don’t know what people want
      • Graph of %content vs. the visits to those pages
      • "Love your loyalists"


    • Rules based engines

      • Click Equations -
        • created a portfolio of best practices and create the rules for you
        • Took the collective intelligence about web analytics
        • Tells you what is broken about your data - and doesn’t let you forget
        • Has good help
        • Can also write your own rules
        • Allows more of your group to be data driven and not


      • Ninjas are better than squirrels


    • Analytics Becomes Intelligent!

      • Have become more and more efficient about puking data, it’s frustrating
      • Need to improve the
      • Donald Rumsfeld quote: known knowns, known unknowns, unknown unknowns (really bad at the latter)
        • the ones we don’t known we don’t know


      • With more an more data this becomes a worse and worse data
      • Puking data is not good enough
      • Control limits ...

        • + Holt-Winter
        • + Possible Causality Factors
        • = something (find in slides)
        • Goal: Better starting points


      • Google Intelligence - Found something interesting about the data

        • What are all of the alerts possible about the data


      • Reveals interesting things like a post from 2 years ago suddenly getting new traffic from new sources
      • Efficient data puking is not the future
      • Leverage Smart Algorithms


    • Outcomes, Baby!

      • Why Don’t We Get Love?
      • Story: Convincing The Boss...
        • "I am kind of a big deal." -not convincing
        • DYK I get 73k visits a day from 176 countries
        • No conversations on the blog
          • Goal 1: All posts ... if you read two posts, you will be convinced of my greatness
          • Goal 2: about page
          • Goal 3: Speaking engagements
          • Goal 4: subscribers
          • Can quantify the value of all of these goals
            • Has an economic value, whether real dollars or (fake dollars in the story he told)




        • What is success to the people who do not convert?
        • You can quantify the value of people commenting of giving reviews (Amazon sales increase after 20 reviews).
        • Don’t just focus on the macro also focus on the micro


      • Listen. Truly Listen.
      • Fail Faster.

        • Most sites suck because hippos created them
          • Highest paid person’s opinion
          • Disconnected from reality


        • Learn to be wrong. Quickly.
        • You can fail on the web every day or every third day.
        • Obama ’08 sign up page is clean an elegant

          • Used Google web optimizer
            • Tried different button labels
            • Tried different images/videos
            • Resulting in a 40.6% conversion improvement


          • Analytics team was 4 people


        • "These kids" raised $500M in 9 months


      • Brand. Shmerand!

        • An auto company wanted to buy keywords for not their own brand because they are already number 1.
        • But after running an experiement, it turns out that buying the brand keywords increased traffic by a large scale.


      • and one more thing...

        • Decision making changes
        • A lot of data coming in and a lot of people who will look at it
        • Funnel it down and create reports
        • As more and more data comes in you start hiring more and more people to manage and report and use that data
          • Problem is that it is not web scalable


        • The data warehouses can get unwieldy and failure prone.
        • What scales on the web

          • Lot’s of data and still have people who need to make decisions
          • Need to move the people closer to the data
          • Hire people who are data literate - data empowered people with a certain level of self help


        • Massive success how?
        • 10/90 rule

          • invest $10 on tools
          • invest $90 on people






    • Death to: (see the slides)


  • Extract yourself from the process. Here is external data, one from our competitors and one from our customers.
  • Web analytics became meaningful when analytics could be tracked via javascript tags

    • When the power shift went from webmasters keeping the site up to the people making the site relevant


  • What is missing in the tools?

    • Only scratching the surface of making the tools smarter
    • Need to empower a certain type of organization of data
    • Everything you learn in one quarter becomes obsolete in the next quarter
    • If you have campaigns on YouTube, ads, twitter ... there is no way to track everything; it’s no longer about the page


  • Are process improvements happening?

    • Yes.
    • On the web "move fast, think smart" but not "think smart, move fast"


  • Are there "click patterns" emerging

    • Yes, in things like shopping chart checkout or structured wizard flows
    • In vast majority of cases this does not occurr
      • Root cause: things are moving to fast
      • Root cause: we at companies don’t know what we are doing