Black swans
tail risk · black swan event · unknown unknowns
A rare event that was not in anyone's model, has very large consequences, and is explained confidently afterwards as though it had been foreseeable. The third property is the one that keeps the mistake repeating.
In practice
The thing that nearly ended the business was not on any risk register, and within a month everyone agreed it had been obvious.
The common mistake
Trying to predict them. The defining feature is that they are outside the model, so the response has to be about exposure rather than forecasting.
Europeans held that all swans were white, on the strongest evidence available: every swan anyone had seen was white. One observation in Australia ended it.
Nassim Taleb used the image for events with three properties together.
Outside the model. Nothing in prior experience pointed to it, so it was not in the range being planned for.
Very large consequences. It dominates the outcome. A few of these determine most of what happens over a long period, in markets and in businesses.
Retrospectively explained. After the fact a clear causal story is available and widely believed, which produces the impression that it could have been foreseen and that the next one will be.
The third property is the interesting one, because it is what prevents anybody learning from the first two.
Why prediction is the wrong response
Being outside the model is definitional. A risk you have listed is not one of these, and improving a forecast can only improve it about the things it already contains.
So the useful work is entirely about exposure rather than about forecasting, which is a different activity and is unsatisfying because it produces no interesting predictions.
Survive the worst case first. Whatever the probability, make sure a single event cannot end you. Concentration is the usual killer — one client, one channel, one platform, one key person. This is the argument under client concentration and it is not about likelihood.
Prefer bounded losses. Anything with an uncapped downside will eventually meet the event that finds it. That is the same convexity point as antifragility, from the loss side.
Keep reserves that look excessive. A buffer sized for the normal case is sized for the wrong case, by construction.
Stay exposed to positive surprises. The asymmetry runs both ways. Many small bets with capped cost keep you in the path of the good version, which arrives as unpredictably.
The one to watch for in yourself
The retrospective story is the part that does damage, because it is so convincing.
After an event, an explanation assembles quickly, everyone agrees on it, and the lesson drawn is specific — guard against this. The next event is a different one. A risk register that grows by one line after each incident is documenting history rather than preparing for anything, which is availability bias with a process wrapped around it.
The transferable lesson from any such event is almost never its content. It is a fact about your exposure: you were concentrated somewhere, or you had no reserve, or a single dependency could stop everything. Those are the things that will be true again when the next unforeseeable thing turns out to be something else entirely.
Concept web
Open the full webQuestions
What is a black swan event?
A rare event that was outside anyone's model, has very large consequences, and is explained confidently after the fact as though it had been predictable. Nassim Taleb popularized the term in 2007.
Can black swan events be predicted?
No, by definition — being outside the model is what makes one. The useful response is managing exposure: surviving the worst case, avoiding uncapped losses, holding reserves and staying open to positive surprises.
What is the lesson from a black swan event?
Almost never its specific content, since the next one will differ. The transferable lesson is about exposure — a concentration, a missing reserve, or a single dependency that could stop everything.