In an earlier post about forecasting, I mentioned the work by Nassim Taleb on the concept of black swan. In his remarkable book, “The Black Swan”, Taleb describes at length the characteristics of environments that can be subject to black swans (unforeseeable, high-impact events).
When we make a forecast, we usually explicitly or implicitly base it on an assumption of continuity in a statistical series. For example, a company building its sales forecast for next year considers past sales, estimates a trend based on these sales, makes some adjustments based on current circumstances and then generates a sales forecast. The hypothesis (or rather assumption, as it is rarely explicit) in this process is that each additional year is not fundamentally different from the previous years. In other words, the distribution of possible values for next year’s sales is Gaussian (or “normal”): the probability that sales are the same is very high; the probability of an extreme variation (doubling or dropping to zero) is very low. In fact, the higher the envisaged variation, the lower the probability that such variation will occur. As a result, it is reasonable to discard extreme values in the forecasts: no marketing director is working on an assumption of sales dropping to zero.
Now, the assumption that a Gaussian-shaped curve’s fit with a potential distribution of outcomes will be the best fit is just that: an assumption. It is based simply on observation of the past. Never before have our sales dropped by 20%, 50% let alone 100%. 10, 20 or 30 years of data can confirm this (observation of the past on a large number of data). But this is only an observation of the past, not a law of physics. Now, if we reason theoretically, not historically, on sales trends, we must recognize that there are many situations in which sales can vary widely. A sudden boycott of our products, for example (Danish dairy products in the Middle East after the Muhammad cartoons), a tidal wave in Japan, which deprives us of an essential supplier, a technological breakthrough that makes our products obsolete (NCR in 1971), the collapse of the Euro, etc. Suddenly deprived of oxygen, our sales are collapsing.
This is the black swan. The reason is simple: sales, like many statistical series, do not follow a Gaussian distribution. The probability of a large variation may be relatively low, but the reality is that in fact it cannot be calculated, because the distribution is unknown and cannot be estimated (this is what economist Frank Knight calls true uncertainty). We can thus be in a year in which the extreme value radically changes the historical distribution. We are in the domain of “fat tails”, ie unlike normally distributed series, high values can have a high probability of occurring. For example, in finance, Taleb noted that 99% of the total variation in the value of portfolios of derivatives between 1988 and 2008 (20 years) occurred in one day, when the European Monetary System collapsed in 1992. Taleb calls environments prone to black swans “Extremistan”.
Extremistan is the area in which a value can change dramatically the sheer scale of the distribution (mean, standard deviation). For example, if you take 10 people and calculate the average of their wealth, it is very likely that if you add a 11th person, the new average will be very different. If you add Bill Gates, a billionaire, the average literally explodes to the point that the other values become negligible, and the standard deviation also explodes. If you do the test at the Davos summit, the standard deviation will be lower. If, on the contrary, you measure the average height of your subjects, adding a new subject will only change slightly the calculated average. The height, in contrast to the wealth, belongs to Mediocristan, the universe of small variations of standard deviation.
Taleb gives another example to explain the phenomenon, that of the turkey. Every day, the farmer takes care of the turkey: he feeds it, pampers it, and ensures its good health. For the turkey, everything is getting better by the day. Every day confirms the feeling that the farmer wants the good of the turkey, and this feeling is based on a statistical series that is longer and longer, with a clear trend: a steady improvement in the variable “weight”, which equals “living conditions”. Tomorrow will be better than today, which is invariably confirmed the next day. One day in late November, the situation changes radically, and the variable “living conditions” drops to zero, completely upsetting the trend of the series. Black swan for the turkey, Thanksgiving has arrived. This example may be funny (except for the turkey) unless when you realize that most of financial experts think just like turkeys. The U.S. real estate has always increased in value, so it will do forever. The Tunisian regime has always been stable, so it always will be. We use Gaussian reasoning in environments that are not Gaussian. We believe we live in Mediocristan while in fact we live mostly in Extremistan.
But Extremistan escapes prediction, even though black swans do not happen that often. So what to do? One approach recommended by Taleb is to minimize the consequences of black swans: if you cannot avoid prediction, maybe you can prepare for their occurrence, and take steps to reduce their impact, should they occur. Raise a Nuclear Power Plant enough above sea level if it is in the area of tidal waves, don’t have a single plant, but two in different geographic areas, etc.
Another approach, only partially helpful however, was mentioned by my colleague Milo in an earlier post on Gresham’s law of strategy, is modeling with Monte-Carlo simulation software and employing explicitly non-gaussian curves.
These models show, however, that reducing the impact of disturbances often involves the development of redundancies, which goes against the quest for optimization (which actually means “optimization for a stable environment”) that is required in modern management. It is as such difficult to accept by the leaders that will always be tempted to favor immediate return at the expense of the resilience of the business. More generally, building redundancies will not be enough. If forecasting really is not possible, we must accept and implement strategic tools that are not based on prediction. We call this non predictive strategy, a concept we will develop in future posts on this blog.
One final note: if you find this perspective on strategy, disruption and surprise interesting, why not subscribe to our blog.