Portfolio Modeling

Description

We're making investments into R&D, so why not manage these investments like we would any other portfolio of investments? We design a model that assigns probabilities to the likelihood that any individual investment will succeed or not, and we build a portfolio of investments based on these probabilities so that our outcome becomes 100% assured. So if there's a 50% chance that any one investment will succeed, then we make two investments and have a 100% chance of success.

Weaknesses

It's complete nonsense. I can't help but feeling that this (e.g., risk-adjusted Monte Carlo modeling) hastened the decline of an industry (pharmaceuticals) that was already in decline. This is stochastic+ reasoning at its worst. When challenged, practitioners give specious reasons for their continuing with this approach, for example: it forces a discipline on R&D projects that is otherwise missing; it may be wrong but it allows us to reassure corporate of our progress; Wall Street demands it, etc.

A hint: any activity having any complexity at all, is not amenable to this type of reasoning. If you read a science article and it mentions the Monte Carlo method (i.e., roll of the dice) then the study authors were missing data and resorted statistical reasoning to hide their lack of data. It's a signal to the reader that the authors of the article are just faking it. For example see the critique of: Recent Warming Reverses Long-Term Arctic Cooling

More damaging, if you're the one being measured it's just too easy to game the system. So we saw the rise of pernicious practices such as WIP counting+ and Stage Gates+ in the pharmaceutical industry. There is no conceivable justification for these practices in industrial R&D.

Appropriate Uses None, except perhaps as a signal to management that something is seriously awry in an organization or study that resorts to these measures. Having and tracking a portfolio is fine. Having portfolio modeling is not.
Further Reading