One idea told three ways: a number without its range is a decision waiting to go wrong. The pieces run in sequence, from the result that is real but unpriced, to the win too small to be worth doing, to the whole system set loose at once. Read in order, or drop in anywhere.
A test can be statistically significant and still be the wrong thing to do. The opener separates the one question statistics answer (could this gap be chance?) from the four it can’t: the range of the lift, the size of the downside, the cost to adopt, and the cost to undo. The economics are computed, not assumed.
A winner can be real and still too small to be worth the switch. This adds the second question every dashboard skips, not is B better, but is it better by enough to pay for acting, and shows why the bar that answers it has to be set before you ever look at the result.
Real decisions don’t move one variable at a time. Here the whole system is set loose, leads, close rate, deal size and lift all varying together, and thousands of simulated runs (Monte Carlo) do the arithmetic algebra can’t, ending on the number a multi-year commitment cares about most: how often it loses money.
The other half of the problem: getting the inputs right. Where the first series reasons once you have the numbers, this is about where the numbers come from, the judgement traps that distort them, and the methods that pull an estimate that means what it says (a calibrated one) out of a person.
When the data run out, decisions run on judgement, and judgement runs on shortcuts that fail in predictable ways. A field guide to availability, representativeness, anchoring, the conjunction fallacy, regression to the mean and overconfidence, with worked marketing examples and the arithmetic that catches each one.
Part I names the traps; this names the fixes, the three-point method, defeating the anchor, reference classes, calibration training, and turning three honest numbers into a distribution you can compute with. Outline only for now.
Before the numbers: the shape of the choice itself. The first series prices uncertainty and the second cleans the inputs; this one structures the decision they feed: which criteria count and by how much, and which choices and unknowns arrive in which order.
Options that are better at different things, and the meeting that argues about options when it really disagrees about weights. Swing weighting in plain words, why importance ratings mislead, and the efficiency frontier that keeps money honestly outside the score.
Draw the decision before pricing it. Folding back in plain words, the average and the lottery it hides, and the tree’s quiet gift: a price on what knowing first would be worth.