Work it through: Deciding Under Uncertainty

Decide how to decide before deciding what to do. Cynefin sorts situations by the relationship between cause and effect: some problems reward analysis, some reward expertise, some only reveal themselves through cheap experiments, and a genuine crisis needs stabilising action before assessment. Most of us default to the method we are comfortable with rather than the one the problem needs. When your engineers call something Complicated and your frontline calls it Complex, treat the disagreement as data and run the cheap experiment before commissioning the expensive analysis.

Watch forPeople place problems by professional preference: engineers find Complicated everywhere, agilists find Complex, and executives under pressure find Chaotic.

For the strands that are genuinely uncertain, build two to four futures that differ in structure rather than in degree, and rehearse your strategy inside each one. The question to ask of every scenario is 'what would we wish we had done?', never 'which one do we believe?'. You are sorting your moves into three groups: the robust ones that work everywhere, the options worth pricing, and the bets. The exercise has worked when the room describes the present differently afterwards.

Watch forA good year, a bad year and the budget is sensitivity analysis wearing a costume, and it leaves the official future untouched.

With the futures rehearsed, price the actual choice. A decision tree lays out the sequence of choices and chance events, puts probabilities and payoffs on the branches, and folds the values back to tell you which first move is worth most. Its deeper service is honesty: every assumption sits on a branch where somebody can challenge it, one number at a time. Give the sensitivity analysis equal billing with the answer, because if the recommendation flips inside the honest range of an estimate, the tree is telling you to go and improve that estimate.

Watch forA positive expected value that risks the survival of the firm is still a bad bet, because you will experience one branch, never the average.

After the domains, the scenarios and the tree, the team will feel rigorous, and feeling rigorous is exactly when overconfidence does its best work. So, before commitment, declare the decision a year old and failed outright, and have everyone write down why, independently and in silence. The silence matters; it flattens the hierarchy for twenty minutes and surfaces what junior people knew and would never have volunteered against a senior sponsor. If the confidence in the room drops afterwards, the exercise has worked. That deflation is the product.

Watch forSoftening the frame to 'imagine things might go wrong' collapses the exercise into an ordinary risk discussion and loses the effect entirely.