What is Expected Value Thinking?

In most leadership meetings, the vivid outcome beats the likely one. The disaster somebody remembers, or the win somebody can already picture, carries more weight than the quiet question of what usually happens. Expected value thinking is the discipline of putting the weights back.

Weighing outcomes by likelihood, not vividness

The idea is old and simple: the value of an uncertain choice is each possible outcome weighted by its probability, summed. It descends from the correspondence between Blaise Pascal and Pierre de Fermat in 1654 on the problem of points, dividing the stakes of an unfinished game fairly. The arithmetic has never been the hard part. The hard part is psychological: Amos Tversky and Daniel Kahneman showed that people judge likelihood by ease of recall, the availability heuristic, so the vivid and the recent loom larger than they should. A firm that just survived a bruising fixed-price project will overprice that risk everywhere; a firm riding a big win will underprice it.

Expected value thinking is the discipline of weighing outcomes by their likelihood rather than their vividness, and of judging yourself on the quality of the decision rather than the luck of the result. Those are two halves of one habit. A good decision can lose; a bad one can win. The poker player and decision researcher Annie Duke calls judging by outcome alone resulting, and organisations do it constantly: the unlucky sound call gets punished, the lucky reckless one gets promoted, and the firm slowly teaches itself to decide badly.

Portfolios, not single bets

A services executive does not make one decision. They make streams of them: bids priced, scopes accepted, hires made, rates held or conceded, escalations absorbed or contested. Expected value earns its keep at that level. Any single bet can go wrong; across a portfolio of decisions, the firm that consistently takes positions with favourable odds and prices will outperform the firm that lurches between fear and euphoria, even though it loses individual bets along the way.

Kahneman and Lovallo observed the organisational failure that follows from missing this: managers evaluate each risk in isolation and turn down gambles a portfolio view would happily accept, while the organisation as a whole ends up both timid in its choices and bold in its forecasts. The portfolio frame licenses sensible risk-taking, and it changes review culture. The question after a loss stops being who to blame and becomes whether the decision was right given what was knowable, which is the only question whose answer makes the next decision better.

Where expected value breaks

Expected value has two failure modes, and knowing them is part of the discipline. The first is ruin. The long-run average only arrives if you survive to keep playing. If one credible downside would sink the practice, the firm or the balance sheet, then a bet can be attractive on expected value and still be wrong, because the sequence ends there. The intuition is as old as the mathematics: Daniel Bernoulli argued in 1738, confronting the St Petersburg paradox, that face-value expected winnings mislead, because the worth of an outcome depends on the position of the person receiving it. A loss that is survivable for a large firm is ruin for a small one, and the same bet is not the same bet.

The second failure mode is the genuine one-shot: selling the firm, betting the company on a single platform, an irreversible commitment that will never recur. With no portfolio for the average to play out across, the distribution you would be averaging over never materialises. In both cases the practical rule is the same: cap the downside first, by structure, by phasing, by insurance or by walking away; judge the option by its worst credible case rather than its average case; and only then let expected value rank what remains.

One concrete example

Clearly illustrative, with no customer implied. A consultancy is weighing a large fixed-price transformation bid. Priced honestly, the likely case is a healthy margin, and the weighted average across the scenarios looks attractive. But the blowout case, delays, rework and penalty clauses together, would consume roughly a year of the practice’s profits and tie up its best people. The portfolio logic is clarifying: a firm that takes ten bets shaped like this should expect to eat a blowout somewhere among them. If one blowout is survivable, this is a good family of bets and the firm should take them repeatedly and calmly. If it is not, expected value was the wrong lens for this decision, and the right move is to change the shape of the bet: phase the programme with break points, cap the exposure contractually, or decline. The firm rescopes into phases: same client, same ambition, ruin removed.

Expected value in the ONX vocabulary

A decision layer makes this discipline routine rather than heroic. A Scenario Run prices candidate options against the facts as they stood, with confidence composed from evidence quality rather than asserted, which is likelihood over vividness made mechanical. The rule that the worst constraint decides encodes the ruin lesson: an option is judged by the constraint that clears last, not by its average, because a commitment is not kept on average. And the Outcomes Ledger is the antidote to resulting: decisions are scored against what was recommended and what was decided, so the firm can tell sound judgement that was unlucky from recklessness that got away with it. That separation, applied across every decision the business makes, is much of what Decision Intelligence is for.