Name Your Criteria Before Claude Compares the Options

A decision workflow where you define the criteria that matter — and their relative weight — before asking Claude to compare options against them, rather than asking an open 'which is better?' that lets the framing float.

'Which should I pick?' invites a persuasive answer built on whichever criteria Claude happens to foreground, and you can't tell if it weighed what you actually care about. Fixing the criteria first turns comparison into evaluation: each option gets scored on the same named axes, the trade-offs become visible, and the decision is yours to make from a table you can audit — not a recommendation you have to trust blind.

Tool selection, vendor choice, architectural options, hiring trade-offs — any decision with several viable options and more than one thing you care about.

Decisions dominated by a single factor (just pick the cheapest that works), or genuinely open exploration where you don't yet know what you value. There, discover the criteria first, then compare.

Help me choose between <options>. First, here are my criteria and how much each matters: <criteria with rough weights>. Add any criterion I've obviously missed and flag it as yours.
Then build a comparison: score each option on each criterion with a one-line justification, note where the evidence is thin, and end with the trade-off I'm really deciding — not a single 'winner'.

Choosing a database, you weight operational simplicity over raw performance; Claude's table shows the 'faster' option loses on ops burden, making explicit that you'd be trading team time for benchmarks you don't need — a trade you can now decide deliberately.

Ask Claude to run a pre-mortem on your leading choice: assume you picked it and it went badly — why? Surfacing the failure modes of the front-runner often reveals a criterion you underweighted.

• Asking 'which is best?' and accepting a framing you never chose.
• Letting Claude invent the weights; the weights encode your priorities, not its.
• Treating the scores as objective truth instead of a structured starting point for judgement.