Verified against ChatGPT · 2026-08-11
Build a decision matrix that exposes which criteria you're secretly prioritizing instead of pretending they're all equal
Scores a real business decision against your stated criteria and weights, then explicitly surfaces the cases where the top-scoring option only wins because of how the weights were set — so you catch a rigged-feeling result before you act on it.
The prompt
Ready to copy — highlighted parts are example details you can swap.
You are building a weighted decision matrix for a real choice I need to make. I don't want a matrix that just outputs a winner — I want to see exactly how sensitive that winner is to the weights I chose, because I've been burned before by a matrix that made a decision look more objective than it actually was. DECISION TO MAKE Which of two office lease options to sign for the next 3 years. OPTIONS ON THE TABLE Option A: downtown, higher rent, better transit access. Option B: suburban, lower rent, requires a shuttle. CRITERIA THAT MATTER Monthly cost, employee commute time, room to grow headcount, lease flexibility. MY INITIAL WEIGHTING (if I have one) Cost matters most, then room to grow, then commute, then flexibility. First, if I didn't give you initial weights, ask me to rank the criteria roughly rather than assuming an even split — an even split across criteria is itself a real choice with consequences, not a neutral default, so don't apply it silently. Score every option against every criterion on a consistent scale, and for any score that's genuinely a guess rather than something backed by real information I gave you, mark it clearly as an assumption rather than presenting it with the same confidence as a score based on stated facts. Calculate the weighted total for each option. Then do the actual valuable part: identify how much the top-scoring option's weighted total would need to shift, or how much a single criterion's weight would need to change, before a different option would win — if the answer is "barely at all," say so plainly, because that means the decision is more finely balanced than the single winning number suggests and shouldn't be treated as a clear-cut answer. If one criterion is doing almost all the work in separating the top two options, name that criterion explicitly, since that's usually the real crux of the decision, not the aggregate score. WHAT NOT TO DO Do not present the winning option as an obviously correct answer if the sensitivity analysis shows it's close. Do not invent criteria I didn't ask for, even if they seem relevant — ask if I want to add them instead. OUTPUT FORMAT The scored matrix as a table (options x criteria, weighted total per option). Then a short sensitivity section: how fragile the top result is, and which single criterion or weight change would flip it.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
A standard decision matrix prompt produces a single confident-looking winner, and the danger is that a weighted score table has the visual authority of objective math even when the weights themselves were arbitrary or unexamined — GPT-5.1 will happily compute a clean weighted sum from whatever numbers it's given without flagging that the inputs, not just the arithmetic, are where the real judgment call lives. Explicitly instructing it to treat an even weighting as a real, consequential choice rather than a neutral default addresses a specific, common mistake: people who don't provide weights usually get an implicit equal split, which quietly encodes the assumption that every criterion matters the same amount, and that assumption is rarely true and rarely examined once a matrix produces a tidy-looking table. The sensitivity analysis is the mechanistically important addition, because it's the part a plain decision matrix never does on its own: rather than stopping at "Option A wins with 8.4 vs 7.9," asking how much a weight would need to shift before the winner changes reframes a close numeric margin as what it actually is — a genuinely uncertain decision — rather than letting a 0.5-point gap masquerade as a clear verdict. Requiring assumption-flagging on any guessed score matters because a matrix mixes fact-based and guessed inputs at the same visual weight by default; without an explicit flag, a confidently-stated made-up score for "lease flexibility" looks identical in the table to a score backed by an actual quoted lease term, and the person reading the output has no way to tell which numbers deserve scrutiny before the decision is finalized.
What you get back
| Option | Cost (40%) | Growth room (30%) | Commute (20%) | Flexibility (10%) | Weighted total | |---|---|---|---|---|---| | A (downtown) | 6 | 8 | 9 | 5* | 7.1 | | B (suburban) | 9 | 6 | 4 | 7* | 6.9 | (*flexibility scores are assumptions — no lease terms provided)
Sensitivity: the gap is 0.2 points — shifting cost weight down by just 5% flips the winner to B. This is a close call, not a clear one; cost weighting is the real crux.
Verified against
ChatGPT GPT-5.1 · 2026-08-11
Changelog
- 2026-08-11 — Initial publish, verified against ChatGPT GPT-5.1.
Need this built into your business?
If a prompt isn't enough — custom software, built and maintained for you — that's Scult's day job.
EXPLORE CUSTOM SOFTWARE
