I have a personal interest in decision making that started from an MBA course on decision analysis. It made me fascinated with the idea of advanced techniques like decision trees with conditional probability versus making decisions with our intuition alone. One of the more interesting things that I have run across that I want to share with you includes the idea that the human mind is a programmed coincidence machine. That is, because of how we have evolved, we are set up to notice unusual cases. For example, when lightening strikes a log we imagine fire. However, commonly, there is not fire when lightening strikes. In short, rather than notice the everyday, mundane, central tendency of a set of occurrences, we recognize the exception to the rule. This is very normal and it is just how, some say, we are built. There are entire books written about intuition versus more explicit decision making. Decision making tools can help us move beyond how we evolved and how we simply react. Here we explore one tool of explicit decision making and its far-reaching effects.
Rigorous decision making has several advantages, including the fact that it is more easily taught, can make our assumptions in decision making very clear, and can force us to be reproducible in our decision making. It also allows us to demonstrate how we arrived at a decision very clearly. One of the most useful tools I have found to achieve these ends is the decision tree.
The decision tree utilizes conditional probability to demonstrate the expected payoff for each possible occurrence in a scenario. Above, I have included a sample from one of our team’s projects. The question is should we invest or not invest in a given startup. You can see, as we work from left to right, we have several different scenarios outlined. Each scenario has an expected payout and a probability of that expected payout occurring. We are able to multiply the expected payout from each branch, times the probability of it occurring plus at each node the similar risk adjusted probability from the other branch. This is called rolling back the decision tree. Eventually, we will obtain an expected payoff from each branch. Here we weigh the expected payoff from investing in a company versus the expected payoff for a more passive income type investment such as investing in mutual funds or a buy and hold strategy in equity based securities. The specifics of what the decision is are not as important as the fact that we can frame it this way and do the best we can to find high quality probabilities for each event occurring. The returns from the stock market are fairly well known, and, by the way, don’t bank on the returns we saw in 2013 continuing. In any event, we can plug in high-quality probabilities to the equity back securities branch (invest in the stock market branch) fairly directly.
Calculating the expected payout from the investment in the new company strategy is more challenging. We are using a home run, base hit, strike out type mentality and nickname each of our branches in that way. Over time, we have come to describe a home run return as one with a 9 x MOI (Multiple Of Investment) return. We are very conservative in assigning our probabilities such that we can make the highest quality decisions. To be conservative, we place this as a low-probability event on the decision tree.
Another useful property to emerge from this is that we are able to examine our assumptions and the probabilities we use in terms of how they affect the model. This is called a sensitivity analysis. We can vary the probabilities and determine if changing the probabilities changes our ultimate answer. Interestingly, sometimes in very complex decision trees, changing our underlying probabilities does not effect the outcome owing to the magnitude of the expected payout, or other probabilities, etc. This is always enlightening and shows us that no matter whether we agree on a certain probability our end decision does not change. This has always been personally fascinating to me and has lead our angel team to what we feel, overall, to be higher quality investment decisions. I invite you to read more about conditional probability along with decision trees, and the progenitor of many of these techniques: John Nash (who received the Noble in part for his work on probabilities in Economics). Consider watching the hollywood movie version of Nash’s life, A Beautiful Mind, if you have some time.
I think you will find the decision tree very useful as has our team. A recent venture capital course, Venture Capital 101, given on Coursera.org highlighted how Ulu ventures and other venture capitalists use similar techniques in their vetting of high quality deals and of decision making in an invest or don’t invest type scenario. Clearly this type of work would not play well on TV and is not readily performed for investment related TV shows like Shark Tank. It would be very challenging to explain this in a way that is palatable to a broad audience. However, conditional probability diagrams and decision trees such as this are very useful to obtain high quality decisions in investment arenas, healthcare, and many other endeavors. Decision making tools can help us move beyond our old mammalian brain and make higher quality decisions.
Speaking of Shark Tank, our next blog entry will include some thoughts on Shark Tank and its positives for innovators as well as entrepreneurs.