It’s A People Question

Published On September 14, 2009 | By mbalogh | Blog, Uncategorized

Recently NetFlix found a winner to their long standing, million dollar, contest designed to inspire freelance teams toward creating new algorithms which would "…substantially improve the accuracy of predictions about how much someone is going to love a movie based on their preferences."  The goal was to improve on NetFlix’s existing algorithm by 10%.  Most teams approached this as a math problem.  One guy in a garage, however, saw it as a people problem and made some significant advances.

For those of you not familiar with NetFlix or the contest, it works like this.  Users log on to the NetFlix website to rent movies.  While on the site they have the option to rank movies using a star system of one to five stars or "not interested".  This information, along with their rental history and a few other bits, is then fed into an application designed to recommend movies the user might like.  The contest concept: give the users better recommendations and they’ll rent more movies and be more satisfied.

Once the contest was made public it didn’t take long for teams to start registering.  A $1 million grand prize can be quite a driver.  Relatively quickly entrants began submitting possible solutions and their results were pretty good.  But then things leveled out.  For about three years.  Then something interesting happened.  Teams began to work together.  One team in particular, "just a guy in a garage", had an interesting approach.

Until recently the identity of team "just a guy in a garage" was unknown but his contest accomplishments, however, were known quite well.  Where other teams were analytical in their approach, he had a background is psychology.  "It’s a people problem", he said, and began looking into how people used the tools NetFlix provided.  What he discovered was several classifications of users.  Some never gave a 5-star rating.  Some never gave a 1-star.  By classifying the people by behavior first, and then running the recommendation algorithm he was able to improve the results quite substantially.

So what does this mean to you?  Everything.  One client recently said to me, "our goal is to get 10-stars on all our service ratings."  Interesting.  What if your clients don’t give 10-stars?  Or what if they rate high?  Wouldn’t it help to know this?  Because knowing the behavior of your customers is knowing how to interpret data and, don’t forget, lack of feedback.

The same thing is true in marketing.  If, for instance, you sent an email with a very attractive offer but saw very low results you’d probably wonder why.  Once you realize it’s July 4th, it becomes clear that a lot of people are probably not checking email.  But then you remember the email was released in the United Kingdom.  And so on.  See how behavior effects outcome?

This is marketing behavior first.  This is Markting 2.0.

Like this Article? Share it!

About The Author

Comments are closed.