This is the Electi review. Today, sensitivity analysis, the practice that separates a model you trust from one you merely hope holds up. Here is the core argument. A forecast is not a decision. It is an assumption dressed up as a number. Every business model rests on inputs price, cost, demand, retention, connected by assumptions that the team rarely challenges. Sensitivity analysis is the discipline of challenging them deliberately. You change one input at a time, or several simultaneously, and you watch what happens to the output. If the result barely moves, your model is solid. If it swings sharply, you have found a fragile assumption, and that is the most valuable thing you can find before committing to a plan. The article identifies three core methods. First, OAT, one factor at a time. The simplest approach. Change one variable, hold everything else fixed, observe the output. Fast, readable, good for executive rooms. Second, global analysis and Monte Carlo simulation, where multiple variables shift within defined ranges simultaneously, producing a distribution of outcomes rather than a single number. Third, sensitivity indices, which quantify the relative weight of each input so teams know where to focus monitoring effort. One finding from the article deserves attention. In the financial sector, a 10% increase in return on invested capital and gross margin can shift enterprise value by more than 20% without growth, and by more than 40% when growth is present. That asymmetry is not a footnote. It tells a CFO exactly which two levers to watch above all others. The article also introduces what it calls decision reversal, the threshold at which a favorable outcome flips unfavorable. This is not a theoretical concept. It is the precise moment a budget plan breaks, a project stops making sense, or a pricing assumption collapses. Knowing that threshold in advance is the difference between a managed risk and a surprise. The practical steps are clear. Clean your data, document your assumptions, define plausible variation ranges, choose your method, and present results as three things the sensitive variables, the size of the effect, and the threshold that changes the decision. A model becomes credible not when it looks accurate, but when you can explain where it is fragile. That is the standard sensitivity analysis holds every forecast to. That's the review.