We owe a debt to Nassim Taleb for memorably encapsulating the demerits of predicting by extrapolating trends as “The Turkey Problem,” and now seems the moment to reiterate it:
Imagine you are a turkey. Every day someone comes to feed you. Every day you get bigger. Your portion sizes get bigger too, brought by a nice man at regular intervals. You extrapolate the trend and you confidently predict a bigger you, with more to eat. Regularly too.
But what happens is … Thanksgiving. Or Christmas
Taleb, N., The Fourth Quadrant: a Map of the Limits of Statistics, Edge Foundation, September 2008
The hard reality for those who predict the future by extrapolating trends (and those gullible enough to believe them) is that even if our turkey had excellent data points (carefully observed and accurately recorded in, for example, a time series analysis) and, moreover, even if our turkey was a mathematically sophisticated — not merely simply projecting trends, but applying all the latest modeling techniques, from moving averages to compound regression — he is still going to be wrong about the future. Dead wrong.
All the data analysis in the world, all the fancy computer software, all the consulting time paid for, and he is still a dead duck.
Ouch. The lesson: there may be (or, vexingly, may not be) something outside the trend, a framing condition, which where it does exist is invisible within the trend projector’s mental model. The only way to get a view of the future that is “robust to Thanksgiving” is (a) to question assumed framing conditions, for example through properly done scenarios, and (b) to hold a view of the future which assumes fundamental ‘game-changing’ surprises can and will occur.
If, as they say, “the trend is your friend” it is assuredly only your fair-weather friend.
