Decision Matrix Calculator

This calculator scores two options across four weighted criteria. The four raw weights are first normalized to fractions that sum to 1 (lines 24-28), then each option's Weighted Score is the sum of its per-criterion scores times those normalized weights (lines 31-36) — Option A's score never reads any of Option B's four criterion scores, and vice versa, since the two formulas are computed independently. At the defaults, Criterion 1's weight (30, the largest of the four) is normalized to 0.30, the biggest share of any criterion — but that alone isn't why Option A — Criterion 1 Score dominates Option A's Weighted Score. Nudging any one score by 10% moves the total by that nudge times the criterion's normalized weight, so it's the score × weight product that decides which input dominates, not the weight alone. At the defaults, a 10% nudge to Criterion 1's score (7) moves the total by 0.7 × 0.30 = 0.21, versus 0.8 × 0.25 = 0.20 for Criterion 3's score (8) — Criterion 1 wins, but only by about 5%, because its score is also large, not just its weight.

The file's own Option B claim proves the point: Criterion 1 carries Option B's largest weight (0.30) too, yet Option B — Criterion 2 Score dominates Option B's Weighted Score instead, because Criterion 2's score (8) times its weight (0.25) beats Criterion 1's score (5) times its larger weight (0.30) once both are multiplied out. Because every normalized weight is non-negative, raising any criterion score for an option, holding the others fixed, always raises that option's own Weighted Score — this holds regardless of which criterion or which option. Winner (line 43) and Margin (line 46) are derived by comparing the two Weighted Scores directly; Winner in particular is a discrete 1/2/0 flag rather than a smooth function, so it can flip from a very small change in either total once the two scores are close. Sensitivity (line 49-51) estimates how much the largest single weight would need to shift to flip the outcome, using Margin divided by the largest normalized weight — a rough approximation, not an exact tipping-point calculation, since shifting one weight also changes how the other three renormalize. This calculator does not model correlation between criteria and does not validate that the four weights entered are individually meaningful beyond auto-normalizing whatever is entered.