# If you are going to say no to almost everything, be logical about it.

Warren Buffet is probably the most famous investor in the world, and maybe the most successful. The 'secret to his success' (as often quoted):

The difference between successful people and really successful people is that really successful people say no to almost everything.

I will add to Warren's quote:

Say no to almost everything and be logical about it, so when you do say yes, people believe you.

Perhaps the most common decision analysis problem I come across, is to justify the purchase of data to help bring clarity to a decision. The most common form of analysis is *Value of Information. *But what is value of information, how do we determine it, and how can be used to say no, and occasionally say yes to an investment. The logic of a value of information analysis goes something like this:

What is the decision in play?

What are my options?

Why am I finding it difficult to choose today?

If it is difficult to choose, there must an uncertainty or uncertainties I am considering.

What would I do if I had to choose today?

What information could I purchase to help make the decision?

What is the value of the decision if I choose today.

If I had the information, what would the value of the decision be.

The difference between the value of the decision with and without data is value of information.

I could go into a long explanation of how to calculate value of information, but to save a very long blog, for those unfamiliar with the analysis of value of information, or who need a refresher, please refer to:

https://en.wikipedia.org/wiki/Value_of_information

https://www.lesswrong.com/posts/vADtvr9iDeYsCDfxd/value-of-information-four-examples

Recently, in preparation for a lecture, I have been considering whether or not to acquire a piece of information to choose between two different strategies for completing and producing an oil well. I went through the normal process of testing the value by considering the decision with and without the information.

My initial analysis I can up with an incremental value of 1.45 million dollars on a base value of 38 million dollars. A significant improvement on value. In theory, I could spend up to 1.45 million dollars and still be ahead of the game. So all good. Go for gold!

Then I posed some further questions:

If we choose to acquire the data, how reliable is the data? In other words if the data tells me 'A' what is the chance it could in reality be 'B'?

How certain are we about the economics of the project?

The experts answer:

I expect the data to be between 60% and 80% reliable.

Due to fiscal and cost uncertainty the economics are plus 10% and minus 20%, so the economic outcome could be 10% better or 20% worse than our forecast.

Before I go on to my conclusion, I should state that data reliability and updating the probability of an outcome is often counter-intuitive and requires a little bit of math, referred to as conditional probability, or Bayesian updating. The complexity is that the revised probability depends on not only on the reliability of information but also the probability of thing you are trying to detect occurring in nature.

Running a quick Monte Carlo simulation revealed the following chart of value across all possible outcomes for the data and the economics:

The green curve shows the combined sensitivity to data reliability and economic uncertainty. The orange dot shows the initial value of information calculation, assuming both perfect information and a single deterministic economic outcome.

A few things can be deduced from this chart:

There is a 15% chance of realising an incremental value of information of 1.45 million dollars or more.

However, there is a 70% chance of eroding value due to:

Potential data unreliability.

Potential economic downside.

The logical conclusion of this analysis is to say no to additional expenditure on data, unless you can be more confident about the data reliability and show some improvement in the economic downside.

So while I agree with Warren, who is a hell of a lot richer than I am, I would add: *be logical in saying no, so when you do say yes, people believe you.*

Last word to Clint Eastwood:

*While based on a real example, the data published here is generic and an artificial case study. *