Why the most useful question is almost always the backwards one.

During the Second World War, the American military had a problem worth solving. Bombers were being lost over Europe, and armour plating was heavy, so you could not simply cover the whole aircraft. You had to choose where to put it.

The obvious approach was to gather data. So they examined the planes returning from missions and mapped where the bullet holes were. The damage clustered on the wings and the fuselage. The conclusion looked straightforward: reinforce where the planes are getting hit.

A statistician named Abraham Wald, working with a research group at Columbia, pointed out that this was exactly backwards.

You are only looking at the planes that came back.

The aircraft in front of them were, by definition, a sample of survivors. The distribution of holes on a returning bomber was not a map of where bombers get hit. It was a map of where a bomber can be hit and still fly home. The places with no holes on the returning planes were the places where damage was fatal, because those aircraft were at the bottom of the Channel and nobody was measuring them.

Armour the parts that look undamaged.

I keep coming back to this story, not because it is a neat piece of wartime trivia, but because it is the cleanest demonstration I know of a habit that keeps turning out to be more useful than the one it replaces.

The habit

Most thinking runs forwards. How do I succeed? What should I do? What is the plan?

Inversion runs the other way. How would I guarantee failure here? What would have to be true for this to go badly? What am I not seeing because of how I gathered the evidence?

Then you avoid those things.

It sounds like a rhetorical trick and it is not. The backwards question is genuinely easier to answer, and the reason is worth understanding.

Failure is specific. Success is vague.

Ask me how to build a business that succeeds and I will give you generalities, because there are countless paths and most of them depend on things nobody controls. Ask me how to guarantee a business fails and I can produce a concrete, actionable list in about ninety seconds. Run out of cash. Build something nobody asked for. Fall out with your cofounder and have no agreement in writing. Ignore the numbers because they are depressing.

That list is short, specific, and every item is preventable. The forwards list is none of those things.

Where the phrase comes from

The usual attribution is to the mathematician Carl Gustav Jacob Jacobi, who is said to have advised man muss immer umkehren, one must always invert, meaning that hard problems often become tractable when approached from the reverse direction.

Small honesty note. That quotation is everywhere and sourced almost nowhere. I could not trace it to a specific text. It may well be genuine, or it may be one of those lines that becomes attributed through repetition. Use the idea, hold the provenance loosely.

Charlie Munger is the reason most people have heard of it. His formulation was that it is not enough to want to be intelligent, you must also work hard at not being reliably stupid, and he was fond of a line about wanting to know where he was going to die so he could simply never go there.

Munger's point was not that avoiding stupidity is a substitute for cleverness. It is that the returns are lopsided. Being slightly cleverer than average is difficult and pays modestly. Avoiding the small set of catastrophic errors that ruin people is easier and pays enormously.

The most useful version of this: the pre-mortem

If you take one practical technique from this, take this one. It is from the psychologist Gary Klein, and it is deceptively simple.

Before a project starts, gather the people involved and tell them the following:

It is eighteen months from now. The project has failed completely. It was a disaster. Write down why.

Not "what are the risks." That question gets you a polite list of hedged concerns. The pre-mortem asserts the failure as an accomplished fact and asks for the explanation, and this changes what people are willing to say.

The mechanism is social, not analytical. In a normal risk review, voicing a serious doubt marks you as the difficult one, the pessimist, the person who is not on board. The pre-mortem removes that cost, because failure is now the premise rather than your opinion. You are not predicting disaster. You are explaining one that has already, hypothetically, happened.

I have never run one that did not surface something real, and usually something at least one person already knew and had not felt able to say.

Where it already runs everything

Once you look for inversion you find it embedded in most fields that cannot afford to be wrong.

Safety engineering is almost entirely inverted. Failure Mode and Effects Analysis exists to enumerate every way a component can fail, and what happens downstream when it does, before anything is built. Nobody designs an aircraft by asking how it succeeds.

Aviation safety may be the most successful example in existence. The entire discipline is built on exhaustively investigating crashes, and the reason flying is as safe as it is has more to do with a century of studying wreckage than with anything else.

Medicine starts from first, do no harm, which is an instruction to weigh the downside before the upside, and much of good clinical practice is subtraction: stopping a medication rather than adding one.

Investing has whole strategies built on it. Understand how you can be ruined, size positions so that ruin cannot happen, and let the returns take care of themselves.

I wrote a while back about why a hundred percent win rate in trading should worry you rather than reassure you. That argument is pure inversion. Instead of asking why the strategy wins, you ask what would have to be true for it to keep winning every single time, and the answer is usually that there is a loss scenario you have not priced yet and have not met.

Two of my own

Health. I built a six month fat loss protocol and the honest version of it started not with the training but with what makes people fail at this. They pick a body fat target they cannot hold. They pile hard training on top of high life stress and stall out around week five. They optimise the training, which burns a few hundred calories, and not the diet, which is the actual lever. Once I inverted it, the design mostly wrote itself.

Storage. I have an ageing server with one drive showing early warning signs. The forwards instinct is obvious: replace the failing drive immediately. The inverted question is what would have to happen for this to end with everything lost, and the answer is that replacing a drive forces a rebuild, a rebuild hammers every remaining disk at full load for hours, and that sustained stress is precisely when a second marginal drive dies.

So the order matters enormously. Back everything up first, then replace the drive. The forwards instinct produces the exact sequence most likely to destroy the data.

Where it breaks, because it does

I would not trust this article if it did not include this part.

Inversion can become paralysis. The list of ways something can fail is effectively infinite, and it is entirely possible to enumerate failure modes until you have talked yourself out of doing anything at all. Avoiding every conceivable risk is itself a way of guaranteeing nothing happens.

Avoiding failure is not the same as achieving success. Not going bankrupt is necessary and nowhere near sufficient. Inversion is excellent at protecting the downside and says very little about how to actually be good at something. If you only ever ask what could go wrong, you will end up safe and unremarkable.

It is weakest exactly where things are genuinely new. The technique depends on being able to anticipate failure modes, and for genuinely novel work, the interesting failures are the ones nobody has seen yet. Inversion is superb for things that have been done before and much less useful at the frontier.

And there is a temperamental trap. If you are naturally anxious, inversion will feel like insight while functioning as rumination. Cataloguing everything that could go wrong is a real analytical technique and also, for some people, a compulsion wearing a technique's clothes. Worth knowing which one you are doing.

How I actually use it

Three questions, and they take about five minutes.

What would guarantee this fails? Then check whether I am currently doing any of them.

How was this evidence collected, and what did that process exclude? This is the Wald question, and it applies far more widely than people think. Every dataset is a set of survivors of something. Customer surveys miss the people who left. Reviews miss the indifferent. Advice from successful people systematically excludes everyone who did the same things and failed.

What is the ruin case, and how likely is it? Not the bad outcome. The unrecoverable one. Most decisions are reversible and deserve speed. A few are not, and those deserve all the caution you have.

That is it. It will not make you brilliant. It reliably stops you being stupid in the specific ways that cost the most, which on the evidence is the better bargain.

The planes that came back had holes in the wings. The lesson was in the ones that did not come back at all.