Leading Under Uncertainty You Can't Reduce

01 Sep, 2026

Or, how to stop paying good money to be told that the future is still in the future

Somewhere right now, a leadership team is entering month nine of a decision that deserved a fortnight. There have been three offsites. There is a data room. There is a workstream studying the findings of the earlier workstream, which studied the market, which was itself commissioned because the first study was judged inconclusive.

The team is not lazy. These are diligent, expensive people doing the most defensible thing available to them, which is thinking visibly. They have converted a hard question into a project plan, and a project plan into the sensation of progress. And they will keep doing this until either the window closes or a competitor makes the decision for them, at which point the research will be retitled as a lessons-learned exercise.

Their problem is not a shortage of information. Their problem is a shortage of an answer. Those are different deficits, and only one of them has a supplier.

Two entirely different ways of not knowing

Epistemic uncertainty is ignorance about something that is already true. The truth is out there 😉. It is sitting in a database, a filing, a warehouse, or the head of a mid-level employee whom nobody has thought to ask. You do not know it, but knowing is purchasable. Effort converts directly into clarity, which is the most satisfying exchange rate in business.

Aleatory uncertainty is not ignorance at all. It is the world not having happened yet. There is no fact to find, because the outcome is still being generated. No amount of diligence extracts an answer that does not exist, in the same way that no amount of squinting reveals the winner of next season.

The cleanest way to hold the distinction is this. Epistemic uncertainty is a coin that has already landed and is covered by someone's palm. Aleatory uncertainty is a coin still in the air. You can bribe the palm. You cannot bribe the air.

Two caveats, because the distinction is a tool rather than a metaphysics.

First, the categories blur, and philosophers have argued for a century about whether anything is irreducibly random or merely stubbornly hidden. Leaders are not entitled to that argument. What matters is reducibility at your budget and inside your window. A question answerable in eighteen months is aleatory on a six-week clock, and treating it otherwise is not rigour, it is astrology with footnotes.

Second, most real decisions are mixtures. The competent move is not to label the whole decision but to decompose it, then route each part to the correct machinery. Count what is countable. Bet on the rest.

Why capable people misfile the second as the first

Nobody misclassifies uncertainty out of stupidity. They do it because the misclassification pays.

Research is unfalsifiable as a career move. Bad decisions have authors. Delays have committees. If you commission a study and the moment passes, the postmortem records that the situation was complex. If you decide and it fails, the postmortem records your name, in bold, with a slide dedicated to it.

Analysis converts dread into activity. Owning an irreducible choice is emotionally expensive, whereas chartering a workstream feels like adulthood. There is an entire consulting industry whose actual product is not information but relief, sold by the week.

Our schooling rewarded only the reducible kind. Exams have answers. Case studies have answers in the teaching notes. Twenty years of that trains a reflex that more work yields more truth, which is true right up until the moment it becomes the most expensive habit you own.

Fake precision is socially rewarded. No one has ever built a five year model with a 3.2 percent terminal growth assumption and meant the point two. But the decimal signals seriousness, and seriousness gets funded. Meanwhile the person who says "this is roughly a coin flip with a good payoff" is described as lacking rigour, mostly by people who have never priced an option.

And the quiet one. Sometimes the report is not intended to inform the decision at all. It is intended for the record, so that whatever happens, someone can point to a binder. That is not a failure of analysis. That is analysis working perfectly, at a purpose nobody will admit to.

A field diagnostic that takes four minutes

Before approving another study, make the room answer these out loud.

Name the fact. What single piece of information, delivered by courier tomorrow at dawn, would change what we do. Not inform. Change. If nobody in the room can name it in one sentence, you do not have a research problem, you have discomfort with a decision.

Locate the fact. Does that thing exist right now, somewhere, in some form. If it comes into existence only after customers, regulators, or rivals act, then you are not gathering information, you are forecasting behaviour, and you should price your report accordingly, which is to say as entertainment.

Price the delay. Learning costs money and time, and delay costs option value, momentum, and occasionally the whole opportunity. Most organisations invoice the first two and never the third. Put a number on waiting, even a bad one, and half of the analysis paralysis in your company dies of embarrassment.

Check the archives. Has this been studied twice already with the same elegant shrug. Repeated inconclusiveness is information, just not the kind anybody wants. It is the world telling you that the coin is still airborne and your researchers are excellent.

Then switch modes entirely

Once you have diagnosed irreducibility, the whole job changes. You stop asking what will happen and start asking what you can survive, how much you are staking, and how fast you can change your mind. Five moves.

Size the bet. The question is no longer whether you are right but whether being wrong is survivable. Distinguish bruises from ruin with actual arithmetic. The single non-negotiable rule of any repeated game is never to take a wager whose downside removes your ability to take the next one. Growth optimises the average. Survival optimises the path.

Buy reversibility on purpose, and know the price. Lease rather than buy. Pilot in one region. Contract before you hire. Build modular so a wrong bet is a swapped component rather than a rewritten platform. Stage the funding. Take the option instead of the acquisition. Reversibility is the only form of foresight that reliably works, because it does not require being right.

But do not fetishise it. Optionality has premiums, and the leader who buys every option owns nothing but premiums. Some doors should be welded shut, loudly, because commitment produces things that flexibility cannot - focus, credibility, deterrence, and the wholehearted effort of people who know there is no plan B to drift toward. Reversibility is a purchase. Sometimes the correct decision is not to make it.

Set trip wires before you are emotionally invested. Write down, now, in numbers and dates, what would tell you this is not working and what you will do about it. Do it today, because future you will not be a scientist evaluating evidence. Future you will be a defence attorney with a mortgage and a reputation, and that person can explain away anything.

Move up to the portfolio. Randomness has no shape at a sample size of one. It develops a shape across many small, weakly correlated bets. If your strategy consists of a single enormous conviction, you have not made a strategy, you have made a wish with a budget attached.

Let action be the instrument. In aleatory conditions, the loop beats the forecast. You do not learn whether the market wants this by convening a panel. You learn by shipping something small enough that being wrong is affordable and real enough that the answer is not hypothetical. Speed is not recklessness. Speed is how you sample a distribution that refuses to be interviewed.

The part nobody prepares you for, which is saying it out loud

Most leaders diagnose irreducibility privately and then present false confidence publicly, because organisations punish visible uncertainty and reward the decimal point. This is how you end up with strategies nobody believes and forecasts everybody knows are theatre.

The alternative is a small vocabulary shift. Separate confidence in the reasoning from confidence in the outcome. "We are highly confident this is the right bet, and we expect a wide range of results" is a coherent sentence and, coming from a leader, an unexpectedly calming one.

Then say the rest of it. Here is the size of the stake. Here is what we are protecting so that a loss stays a loss rather than becoming a crisis. Here is what would make us fold. Here is the date we look again. That is not hedging. That is the only honest form of executive courage, and it reads as strength to everyone except the person who wanted a number they could quote back to you in a year.

And protect the people who make well sized bets that lose. If you punish outcomes rather than judgement, you will be governed within eighteen months by the most sophisticated report writers in your industry, and none of them will ever decide anything again.

The point

Wisdom here is not knowing more. It is knowing which kind of ignorance you are standing in, and having the discipline to stop shopping when the shop does not stock what you need.

The mature leader is not the one who removed the uncertainty. That was never available. It is the one who stopped pretending removal was on the menu, and instead arranged to be wrong cheaply, quickly, survivably, and on purpose - which, over enough rolls, is indistinguishable from being right.