Build a Weighted Scorecard, Not a Vibe

A comparison method where you turn a decision into a scored matrix: list the criteria that matter, assign each a weight reflecting how much it matters, have Claude score every option against every criterion, and compute a weighted total. The output is a ranking you can inspect, not a paragraph of on-the-one-hand prose.

Unstructured comparisons hide the reasoning: a wall of pros and cons leaves the actual trade-off — how much you're willing to sacrifice on one axis for gains on another — implicit and unexaminable. A weighted scorecard forces that trade-off into the open. Because the weights are yours and visible, you can argue with them; because each score is per-criterion, you can see exactly why one option leads and whether the margin is real or an artefact of one weight. It converts 'trust my judgement' into 'here's the model of the decision', which is both more honest and easier to pressure-test.

Multi-criteria decisions with real stakes and several viable options: choosing a vendor, a tech stack, a hire, a strategy. Anywhere the answer is genuinely 'it depends' and you want to make the 'depends' explicit rather than resolve it by gut feel.

Decisions dominated by a single hard constraint (it's over budget — done) or where quantifying would be false precision on inherently unquantifiable factors. A scorecard is a thinking aid, not a way to launder a gut call into a number.

Help me compare <options> as a weighted scorecard. First propose 5–7 decision criteria and suggest a weight for each (summing to 100) — I'll adjust. Then score each option 1–5 on each criterion, with a one-line justification per score. Compute the weighted total and rank them. Finally, tell me which single weight, if I changed it, would flip the ranking.

Choosing between three project management tools, you weight 'team adoption' at 40, 'integrations' at 30, 'price' at 20, 'reporting' at 10. Claude scores each and the winner isn't the one you'd assumed — it lost on integrations but the modest weight there couldn't overcome its lead on adoption. Seeing that, you agree: adoption really does matter most, and the scorecard just made your own priorities honest.

After the ranking, run a sensitivity check: ask Claude which criteria the result is fragile to and what the ranking would be under a plausibly different weighting. If the same option wins across reasonable weightings, you have a robust decision; if it flips easily, the real question isn't which option — it's which weights you actually believe, and that's where to spend your judgement.

• Setting the weights after seeing the scores, so you tune them until your preferred option wins — reverse-engineering, not deciding.
• Inventing precise scores for factors that are genuinely qualitative, dressing a hunch as arithmetic.
• Reading only the totals and ignoring the per-criterion scores, where the actual trade-off and any close margin live.