I have just finished reading Nassim Nicholas Taleb’s book – The Black Swan: The Impact of the Highly Improbable. NNT (as he calls himself) has some fascinating points and some interesting turns of phrase, though he does rather go on and on and on…. Leaving aside the long-winded somewhat self-absorbed writing style, NNT makes some interesting points that he illustrates well. He discusses the problems with the fact that humans seek validation for what they think (rather than challenging themselves) and why we cannot predict well in many circumstances. He spends a lot of time discussing the problems inherent with the use of a bell curve to predict things where the impact of extreme events really matter. Finally he spends (too little) time on what to do about all this.
I took a few key points away from the book:
- It is better, perhaps, to try and be generally right rather than precisely wrong
Don’t throw out unusual situations just to make your response to the usual look more precise
- Beware of looking for more rules than really exist
- Watch out for a tendency to prepare for “last war”
Many regulations and policies are focused on preventing the last bad thing that happened and too little time is spent worrying about what might happen next
- Rare and consequential events can be much more important than the “normal” stuff in the bell curve
- Some things (that he calls “mediocristan”) are such that the most typical is average and single instances don’t impact the total much
The weight or height of people for instance. These things can be managed with a bell curve mentality
- Others (he calls these “extremistan”) are more winner-takes-all kinds of environments where extreme events matter most
- He makes the point that a thousand days cannot prove you right but one can prove you wrong
There’s more but it’s a very long book and I am not going to attempt to summarize the nuggets spread throughout it here. There were a couple of points with respect to decision management I thought I would make:
- He makes the somewhat tangential point that good news spread out works better than big lumps of it – people are happier with lots of small wins. This could have an impact on how you market to customers in terms of best next action. Don’t try and win over customers all at once, focus on lots of small wins.
- There is a problem of “silent evidence” – a risk of ignoring the customers you never signed. So when modeling it is important to consider the kinds of customers you don’t have (because of how you vetted or attracted potential customers in the past). There are a number of analytic techniques in this area, such as reject inference (inferring the behavior of customers you rejected).
- There is a problem where expertise, data and experience on the part of staff improve their perceived accuracy even when it does not improve actual accuracy! This is a general call for driving your decisions from data and analytics, not from people’s judgment, where you can.
- In general the need for agility is clear in all of this. You cannot possibly model all the things that could matter so being able to change your decisions, and hence your systems and processes, quickly at least enables you to respond quickly to the unexpected.
Originally published on the EDM blog.