GenuineVC David Beisel's Perspective on Digital Change

May 10, 2005

People often mistakenly derive extraneous meaning out of the fact that two things are merely correlated. They will say, “A increases as B increases; if we change A, then that will affect B.”

It’s actually more than a pet peeve, a minor irritation. It’s really flawed logic.

Correlation simply doesn’t imply causation.

Just because two trends are correlated doesn’t mean one causes the other. Often, a third related factor causes both originally identified trends.

For example, just because in the past few years we’ve seen an increase in both search engine spending and the number of SEO/SEM firms, it doesn’t mean that one of these caused the other. Will increasing search engine spending directly increase the number of SEO/SEM firms, or vis-versa? Perhaps the real driver is the number of advertisers who have experimented and enjoyed successful results from these marketing efforts. Or maybe it’s a virtuous cycle in which it’s difficult to truly unmask the true cause.

Along those lines, we often see business plans here at Masthead that claim “because Factor X increases as the customer’s revenue-per-user increases, our product which affects Factor X will increase revenue even more.” Really? Maybe, but often there is a Factor Y which is the real cause of both Factor X and revenue. A company’s product offering should really get to the heart of the issue and the real driver of value.

When people present correlation graphs, I stop paying attention when they don’t know the difference between the two ideas of correlation and causation. This mistake is very dangerous, as it often leads people to chase the wrong goals.

  • Scott

    As a former student of Science I must completely agree with David regarding this issue. Even in the real world of business we quickly (and sometimes painfully) learn that correlation may indeed be necessary; however, it does not provide a confident means of determining causal relationships. Most importantly, correlation should never be mistaken as a substitute to the experimental method and the “scientific process.”

    -Scott

  • Gibson

    Yes, but if you only are able to observe correlation, it can be a good starting point.

    For example, when an organization uncovers a problem and begins assessing the cause, examining correlations can assist as a starting point to uncover the actual cause. If turnover is a big issue for a company, understanding the correlations between increased turnover and other variables can assist in isolating the true causes. True, additional investigation is required to understand the correlations to deterimine whether or not they are in fact causes, but the correlations are able to at least lead the way…

    Potential companies that use correlational analysis to justify their plan can be a starting point, just not the entire answer…

  • Christy

    I’m a scientist, and I agree… but do people in business still really make this mistake often? Seems like we learned it on the first day of general linear models.

  • http://www.myhome.com Lier

    Your essay was very good, in your articles I learned a lot, as someone who just learn English beginners. I'm honored to see such worthy of study of articles. This is the sea of knowledge.replica hermes watches

About Me

  • avatar
  • I am a cofounder and Partner at NextView Ventures, a dedicated seed-stage venture capital firm making investments in internet-enabled startups. Read More »

Coordinates

Subscribe

Rob Cho Go




 RobGo.org

NextView Twitter Stream

51015
  • Lee Hower
     - 1 hour ago
    board deck templates for seed stage companies - our next Growth Guide from @NextViewVC is out http://t.co/5oTNAlt5BR
  • robgo
     - 2 hours ago
    @mattcharris @joshelman funny, I was just thinking about getting that version
  • robgo
     - 2 hours ago
    @joshelman thanks! Just making it easy for you when you join the board of one of our portfolio companies :)
  • robgo
     - 2 hours ago
    @vintage141 thanks!

Search