Balancing Intuition and Analytics in Decision Making #PBLS

October 29, 2009

in Analytics, Strategy


Malcolm Gladwell, Thornton May (author of The New Know: Innovation Powered by Analytics)and Tom Davenport (author of Competing on Analytics, reviewed here) made up a high powered panel for this. Various random comments follow:

  • Healthcare is being used as an example by the panel as an obvious point where analytics and expertise intersect. There is a challenge to creating in experts because they need so many hours – 10,000 as Malcolm mentioned – and time pressure is making harder and harder to get this. When the number of facts you need to know to be an expert (medical specialists are often said to need to known 2,000,000 or more facts) there is a compelling need to have systems be involved. However, the different criteria for success – patients often have different, less precise criteria – mean that people are unlikely to be replaced completely by computers. The empathy of doctors, their "bedside manner" is not going to be replaced by a machine but it should be supported and given context by one.
  • There is a critical difference between experience-based intuition and unaided intuition or "gut feel". Businesses must value the former but what about the latter? When does one trust intuition rather than analytics. The first criteria is how often we have done this before – if we have done it before often then we should use analytics. If it is a new problem, an attempt to be radically different, then intuition is critical. The second is related – how much data do we have about the problem.
  • Financial crises is a crucial test case for analytics – all these bad decisions were made by people who had sophisticated analytics. The analytics led these folks to believe they could manage all elements of risk and the analytics got into the hands of people who did not understand the limitations of the models. This is a critical issue – the executives have to be able to understand the analytics, they must be analytically informed. Similarly there is a limit to how much can be modeled, some risks for instance are just out of the ordinary (Black Swans, as they are known). Modelers must be clear these risks are not in the model. And they must make sure that everyone downstream from them understands these limitations.
  • The tools you can create with analytics, the very useful tools, you must also have an appropriate context for these tools. A regulatory framework, for instance, provides context for a model. There is a skills gap between executives and analytics folks but a larger one between the analytics capabilities of the regulators and those they regulate. Why, for instance, are regulators all lawyers rather than analytics people? No simulations or models were done, for instance, on the impact of the stimulus money. Not enough analytic depth in the regulatory framework.
  • We must understand intuition as the fruit, the outcome, of many years of study and experience. To create people with good intuition we must be willing to have them spend time on decision making. And those who are good decision makers must learn from their mistakes, they must be more proactive in analyzing how their intuition let them down as they can improve it. A willingness to engage in introspection is essential.
  • Human beings are not good decision makers about their finances and it is time to have regulators enforce some ethics on the ability of companies to use analytics to manipulate consumer behavior – companies are getting way smarter than consumers and are using it to manipulate them.
  • It is just as important to be able to tell a good story with data, as an analyst, as it is to be able to do the analytics.

Closing thoughts:

  • Tom: Lots of opportunities, lots of new tools, more understanding of how we make decisions. Time to systematically look at how we make decisions.
  • Thornton: World is digitizing so the cost of experimentation is falling and the opportunities are greater than ever
  • Malcolm: We must remember that failure is critical in experimentation and make it cheap and easy to fail and to minimize the impact of failure.

These guys were fun, but hard to blog. Hope the post is helpful anyway.


Previous post:

Next post: