Verified against ChatGPT · 2026-07-26
Build a weighted decision matrix that can't be gamed by picking favorite-friendly weights
Constructs a weighted-criteria decision matrix for choosing between real options, with an explicit sensitivity check showing whether the recommended winner actually holds up across a reasonable range of weightings, instead of a single static score that quietly depended on one specific weight split.
The prompt
Ready to copy — highlighted parts are example details you can swap.
You are building a weighted decision matrix to help choose between the options below, on the criteria named — and then stress-testing whether the winning option actually holds up, or whether it only wins because of one specific weighting choice. OPTIONS BEING COMPARED Three office lease options: Building A (cheaper, further from transit), Building B (mid-price, on transit line), Building C (most expensive, includes a build-out allowance). DECISION CRITERIA Monthly cost, commute accessibility for the team, and flexibility to expand headcount within the lease term. INITIAL WEIGHT SPLIT Cost 40%, commute accessibility 35%, expansion flexibility 25%. WHAT ACTUALLY MATTERS MOST IF IT COMES DOWN TO A CLOSE CALL If it's genuinely close, commute accessibility wins — a bad commute is the thing most likely to actually cost the team people. MATRIX RULES Score every option against every criterion on a consistent numeric scale, and for each score, state the specific evidence or reasoning behind that number in one line, not just the number alone — a matrix full of unexplained scores invites both sides to unconsciously adjust a score to match a preferred outcome rather than reflect the actual evidence, and a stated one-line justification per score is what makes each one checkable independently. Apply the initial weight split given to compute the first-pass winner, but then run a sensitivity check: identify how much the weights would need to shift before a different option becomes the winner, and name specifically which criterion's weight is most responsible for the current result — if the winning option only wins under a narrow band of weightings and a small, defensible change flips the result, that's a materially different and more fragile answer than one that wins across a wide range of reasonable weightings, and the matrix should say which situation this actually is rather than presenting a single static score as if it were equally decisive either way. If two options score within a small margin of each other after weighting, don't round that up to a confident recommendation — say explicitly that it's a close call and use the stated tiebreaker priority to make the actual call, naming that it's being used specifically because the weighted scores alone didn't clearly separate the two. Do not let a criterion weighted low quietly override one weighted high just because it happens to have a more dramatic score spread between options — the stated weights represent what actually matters, and a criterion's score spread being more visually dramatic is not a reason to let it carry more real influence than its stated weight assigns it. OUTPUT FORMAT 1. The full scoring matrix: Option (rows) by Criterion (columns), each cell as score plus one-line justification. 2. Weighted totals and the first-pass winner. 3. The sensitivity check: how much weights would need to shift to flip the result, and which criterion's weight is most load-bearing. 4. If it was a close call, one paragraph applying the tiebreaker priority explicitly.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
A decision matrix's output is only as trustworthy as its inputs, and an unexplained numeric score is trivially easy to nudge toward a preferred conclusion without anyone necessarily noticing, because a bare number carries no evidence trail a reader could push back on; requiring a one-line justification per score converts each cell from an opinion into a checkable claim, which is what actually lets a reviewer catch a score that doesn't match its own stated reasoning. The sensitivity check targets a specific and common blind spot in weighted scoring exercises: a single weighted total presents one specific weighting choice's output as though it were the answer, when in reality most real decisions have genuine, defensible disagreement about the exact weight split, and a result that only holds under one narrow weighting is a fundamentally weaker basis for a decision than one that's robust across a wide range of reasonable splits. Explicitly flagging a close call and deferring to a pre-stated tiebreaker, rather than rounding a small margin up to a confident recommendation, prevents a specific kind of overclaiming: weighted scores computed to decimal precision create an illusion of exactness that a genuinely close real-world tradeoff doesn't actually have, and presenting a fractional-point margin as a clear winner manufactures false confidence in a distinction the scoring method wasn't actually precise enough to draw. Protecting a low-weighted criterion from being overridden by a more visually dramatic score spread matters because that's a specific and easy-to-miss way a matrix can quietly betray its own stated weights — an unweighted glance at raw scores can read a more dramatic spread on a lower-weighted criterion as more decisive than the math says it should be, and the matrix has to hold the actual arithmetic to the stated weights rather than to which column visually looks more differentiated.
Verified against
ChatGPT GPT-5.1 · 2026-07-26
Changelog
- 2026-07-26 — Initial publish, verified against ChatGPT GPT-5.1.
Building this for real?
This is a free starting point. If you'd rather have what Scult builds built and running for your business, that's Scult's day job.
EXPLORE WHAT SCULT BUILDS
