May 29, 2012

Attribution: What Do We Really Know?

Here's a few thoughts for you.  Please read these quotes:
  • "Mason Crosby kicked a 42 yard field goal as time expired, leading the Green Bay Packers to a 24-21 victory over the New York Giants."
  • "The Dow Jones Industrial Average dropped 124 points, or 1.1%, on reports of a two billion dollar trading loss at JPMorgan Chase."
  • "Our meeting won't start until 10:15am, because Penny is held up in a traffic jam."
  • "My computer is running slow because it is infected with a virus."
  • "Janice, a fifteen year veteran, was promoted to Vice President of E-Commerce because of her outstanding work optimizing our email marketing program during the past year."
Each quote offers some form of attribution, correct?

Let's be honest.  Does Mason Crosby deserve full credit for a win?  What about the other 120 plays in the game, contested by forty-four teammates?  What role did each player, and each play, have in the outcome of the game?

Can anybody prove that a two billion dollar trading loss at one company, by one individual, caused billions of shares traded in nanoseconds by computer algorithms to yield reduced market capitalization at companies not remotely connected to the trading loss?

Can we prove that a traffic jam kept Penny from getting to a meeting on time?  Did Penny leave for the meeting soon enough?  Could Penny have chosen a different route?

Even if a computer is running slowly because of a virus, could the user have done anything different to avoid getting a computer virus?

What role did the other fourteen years play in Janice becoming a Vice President?

In life, we accept ambiguity, don't we?  We don't demand a full accounting for how each project that Janice worked on over fifteen years contributed to her promotion.

But in Marketing, we view this differently?  We somehow think that we can parse every stinking click, every visit to the mall, and every silly tweet into a category that results in the division of a $100 order across the 27 actions that led up to the order.

What do we really know about the value of all 27 actions that led to a $100 order?



  1. Kevin

    An interesting question and one that I am compelled to answer.

    Do we really know the value of all the 27 actions? No

    But do we know that there were 26 actions before a user finally converted on the 27th? Yes.

    Therefore, having this information at our fingertips means we have two choices...
    1. Look at it & try to do something with it
    2. Ignore it and measure eveything on last click

    Personally, I would do the former, if only to get some insight into customer behaviours and see what channels they did use.

    Sure.... the attribution approach is complex, but by clever and detailed use of analytics, we can start to see which channels make the most difference to our bottom line and suitably reward them.

  2. That's an acceptable answer.

    Of course, if we give this problem to 100 experts, we'll get 100 different answers, each equally likely to be right, each equally likely to be wrong. None represents truth. But you'll have 100 experts all arguing about how wrong last-click stuff is, few admitting the errors/biases of their solution.

    And that's the point of this discussion. What do we really know?


Note: Only a member of this blog may post a comment.

The Milkman

In the mockumentary "Unfrosted", Post and Kellog's are in an arms race to create the first unfrosted toaster pastry. Seinfeld&...