In January 2011, the operators of Wivenhoe Dam outside Brisbane had more information than most Deciders ever get. They had live inflow data, Bureau of Meteorology forecasts, and a rulebook telling them when and how much water to release. The city still flooded. Decision-making under uncertainty is the practice of deciding while uncertainty remains by naming the assumptions carrying the decision, judging their significance, and deciding whether what remains is sufficient for the Purpose at stake. Wivenhoe failed not because uncertainty existed, but because the assumptions inside the rulebook had stopped matching the world.
That is the wrong-question problem at the centre of this subject. People ask how to decide when they do not have all the facts. You never have all the facts. You never will. Every live decision rests on assumptions about markets, weather, regulators, suppliers, politics, timing, and human follow-through. The real question is different: which assumptions matter, how much confidence do you have in them, and what would count as enough certainty to act?
I have spent nearly fifty years watching organisations answer that question badly. They commission more analysis, ask for more modelling, and wait for uncertainty to retreat. It does not. What changes, if the process is any good, is the quality of the Decider's judgment about the assumptions doing the real work. That is what the Universal Decision-Making Method is for, and it is the practical question Grant Purdy and Roger Estall kept returning to across boards, regulators, and public bodies.
What decision-making under uncertainty actually means
Decision-making under uncertainty is not a specialist corner of strategy or economics. It is the normal condition of deciding. A pilot deciding whether to divert, a board deciding whether to expand, a founder deciding which market to enter, and a family deciding whether to move all face the same structure. They know some things. They assume others. The outcome depends on both.
Frank Knight drew the classic distinction more than a century ago in Risk, Uncertainty, and Profit: risk is measurable, uncertainty is not. That matters because most organisational machinery still behaves as if uncertainty were simply risk that has not been modelled hard enough yet. The instinct is always to quantify one more thing. The harder question is whether the assumptions carrying the decision are good enough to act on.
The mistake is to imagine that uncertainty is the unusual part and certainty the baseline. It is the other way round. Certainty is local and temporary. Today's fact becomes tomorrow's assumption as soon as time passes or conditions shift. Production costs look factual at the moment a pricing decision is made, then become assumptions about whether those costs will remain stable long enough for the decision to work.
That is why the search term matters. People who ask about a decision under uncertainty are usually not asking for abstract theory. They are asking for a way to decide without pretending the world has become stable first.
Why waiting for all the facts makes it worse
Wivenhoe is the cleanest example because it strips away the excuse that more data would have solved it. The operators had data. What they lacked was a decision process that forced a confrontation with the assumptions inside the operating manual. The Queensland Floods Commission of Inquiry later documented how central those assumptions were, about inflow behaviour, downstream capacity, and the continuing validity of the release rules. Because they were embedded in procedure rather than made explicit, nobody could test them clearly at the moment they mattered most.
The same failure appears in slower-moving settings. A board waits for one more forecast. A project steering committee asks for one more round of review. An executive team assumes that additional diligence is reducing uncertainty when in reality it is merely increasing paperwork. Spyros Makridakis, Robin Hogarth, and Anil Gaba made the same point in Forecasting and uncertainty in the economic and business world: forecasting is crucial to business decisions, yet accurate forecasting is usually not possible in the environments where those decisions matter most. The extra effort can be rational for a while. Then it turns into delay that costs more than the uncertainty it was supposed to reduce.
This is why the usual advice, gather more information until you feel comfortable, is weak. Comfort is not a decision standard. Neither is completeness. The right standard is whether the assumptions driving the choice have been surfaced, tested where they can be tested, and judged honestly where they cannot.
When decision analysis under uncertainty stops exposing those assumptions and starts defending a preferred answer, the model becomes camouflage.
Assumptions are the real object of the decision
The most useful question in the whole method is brutally plain: what are the assumptions we are making here? The moment that question is asked properly, a foggy decision starts to become visible. Options that looked equally plausible separate. Arguments that sounded confident reveal that they were resting on borrowed certainty. Disagreement becomes more productive because people can now point to a specific assumption instead of gesturing at a general unease.
Not every assumption matters equally. That is why Roger Estall and I use a significance matrix with two axes: how much influence the assumption has on the desired outcome, and how confident the Decider is that it will hold. High influence with low confidence is Critical. High influence with high confidence is Important. Low influence with low confidence is Relevant. Low influence with high confidence is Limited.
The matrix does not make the decision for you. It tells you where the next hour of work belongs. A Critical assumption can be dealt with in only three sensible ways: get better information, modify the decision so it depends less on that assumption, or choose a different option entirely. Everything else is drift.
If you want to see what that looks like on a live executive decision rather than in a diagram, the clearest nearby example is how to make a difficult business decision. The point is not to produce a prettier model. It is to compress a vague, over-analysed question into the few assumptions actually carrying the outcome.
Organisations that rely on dashboards and data models to guide strategy still face the same question. Business decision analytics under uncertainty examines where the numbers stop and the Decider's judgment on assumptions begins.
Sufficient certainty, not maximum certainty
The phrase that matters most here is sufficient certainty. Not maximum certainty. Not perfect information. Enough. That is the stopping rule most governance processes lack.
The blood-imports case in Deciding shows why this matters. In trying to reduce the chance of contaminated blood reaching patients, some countries suspended all imports from suspect sources. The result was a blood shortage and increased mortality. Greater certainty on one narrow question damaged the broader Purpose the decision was supposed to serve.
Sufficient certainty is the discipline that prevents this. The Decider asks whether the remaining uncertainty is acceptable given what is at stake, given the significance of the assumptions involved, and given the cost of chasing more certainty. That judgment cannot be outsourced to a formula. It belongs to the Decider because the consequences do too.
When organisations refuse to work this way, the pattern usually turns into analysis paralysis. The analysis continues because nobody has defined what would count as enough.
Deep uncertainty, vulnerability, and context drift
Some decisions are harder because the uncertainty is not just large but structurally unstable. The wider context itself may be changing faster than the organisation can observe. That is where deep uncertainty enters, not as an exotic theory term but as a practical condition in which historical probabilities help less and context monitoring matters more.
Appendix B of the book pushes this further than the symptom hubs do. The point is not only that change can occur. It is that organisations create vulnerability to change through their own earlier decisions. Buildings collapse in earthquakes because someone decided what standards they would be built to. Business models fail because someone decided what would be watched and what could safely be ignored. A decision made under uncertainty never ends when it is approved. It continues through the assumptions it leaves behind.
National postal services show the slower version of the same lesson. Letter volumes declined as email spread, but parcel volumes grew with e-commerce. The organisations that kept treating themselves as letter businesses saw change as a threat arriving from outside. The ones that noticed the context shift early could redesign around a different demand pattern. The decision problem was not whether change existed. It was whether anyone had been watching the right assumptions about how people would communicate and buy. That too is decision-making under uncertainty: not more information, but better attention to what can change.
That is why the uncertainty cluster needs to touch disruption and monitoring directly. Uncertainty is not just what exists before the decision. It is what keeps moving after the decision has been made. When the probabilities themselves are too weak or contested to carry a decision, decision making under deep uncertainty becomes a practical problem, not an academic one.
What this looks like in business decisions
In business, uncertainty often arrives disguised as abundance. A team has forecasts, committee notes, consultant packs, scenario models, and contradictory confident opinions from people with different incentives. The volume creates the impression that the decision is being worked properly, even when nobody has yet stated what the decision actually rests on. If this sounds familiar, the page on decision-making frameworks separates the tools that produce analysis from the method that produces a decision.
Take a company deciding whether to expand fulfilment capacity into a second site. The useful questions are not endless. Will demand hold long enough to justify the lease? Can management absorb the operational complexity? Are labour availability and transport costs stable enough to support the move? Those are business-decision questions under uncertainty. They are manageable once named. They are unmanageable while buried in a larger apparatus that treats every uncertainty as equally important.
Sometimes a Decider can do this work alone. Sometimes they need a more disciplined conversation, which is where decision coaching belongs. The coach does not remove uncertainty and does not take the choice away. The role is to make the reasoning visible enough that the Decider can say, plainly, whether the assumptions are good enough to proceed.
If the distinction still feels abstract, read the decision making under uncertainty example cases. Apollo 13, Blockbuster and Grenfell Tower each show a different failure mode when assumptions stay hidden.
The five steps that make this workable
The method itself is simple. Frame the decision. Develop options. Recognise assumptions. Sufficient certainty. Design monitoring. Those are the same five steps set out on the pocket card and on the main Method page.
What matters here is not memorising the names. It is understanding what they do to uncertainty. Framing stops the Decider from solving the wrong problem. Options stop the process collapsing too early onto one preferred story. Recognising assumptions turns hidden uncertainty into visible statements, whether the assumption arrived through availability bias, an anchored reference point, or a preferred conclusion nobody questioned. Sufficient certainty gives the loop a stopping condition. Monitoring keeps the decision alive after approval instead of pretending the world will stay still.
When the problem itself is tangled, the discipline is the same. Complex problem solving begins by naming the next decision instead of reaching for a grander framework. When that tangle sits inside a strategy conversation, strategic problem solving shows how the same five steps apply.
The method is not rigid. Sully used the same structure in seconds. A board may use it across weeks. The difference is not the method. It is the time available and the speed with which the context is changing. When the clock is genuinely short, the same sequence still applies, it just has to be built before the pressure arrives. That is the argument in decision making under pressure.
A closer look at how each step handles uncertainty in sequence is in the decision-making process under uncertainty.
What to monitor after you decide
A decision under uncertainty is only as good as its monitoring design. If an Important assumption stops holding three months later and nobody is watching for that, the original quality of the decision becomes almost irrelevant. The outcome will drift and the organisation will discover it late.
Good monitoring asks simple questions in advance. What must still be true for this decision to keep making sense? How likely is each thing to change? How quickly would that change matter? How easy would it be to detect? The answers tell the Decider where attention belongs after the decision leaves the room.
That is also why the decision autopsy matters. It is not an exercise in blame. It is a way of reconstructing what was assumed, what later changed, and whether the change should have been visible sooner. Organisations that do this well get better at uncertainty over time because they stop treating each bad outcome as a mystery.
Sufficient certainty is the target at the moment of decision. Monitoring is what stops that certainty from fossilising into wishful thinking afterwards.
Grant Purdy is the co-author, with Roger Estall, of Deciding (2020), and the architect of the Universal Decision-Making Method.
If you have a decision you are working through, the Walk can help.
Start a Walk