# Picking the Prioritisation Framework (Instead of Defaulting to RICE)

Frameworks aren't interchangeable — each encodes assumptions about what you know and what you're optimising. Picking the wrong one produces confident garbage. This is the selection logic.

## The selector

| Your situation | Use | Because |
|---|---|---|
| Comparable initiatives, decent usage data, one primary metric | **RICE** | Its inputs (reach, impact) are estimable; its output ranks like-for-like |
| A release scope to defend under deadline pressure | **MoSCoW** | The question is in/out, not rank; Must/Should forces the cut conversation |
| Deciding what makes an experience *feel* complete vs delightful | **Kano** | Separates hygiene (absence infuriates) from delight (presence differentiates) — invisible to RICE |
| Early product, thin data, competing bets | **ICE** | Honest about estimate quality; fast; doesn't fake precision RICE would |
| Rich discovery data on underserved needs | **Opportunity scoring** | Ranks by importance-vs-satisfaction gap — finds what users can't ask for as features |

## Cross-framework truths

- **Effort is always the most-lied-about input.** Whatever framework: effort estimates come from the people who'd build it, include rollout/support, and use coarse buckets. PM-solo effort guesses are the #1 corruption vector.
- **A framework ranks; it doesn't decide.** Dependencies, strategic sequencing, and commitments legitimately reorder the list — the discipline is writing down WHY you deviated, not never deviating.
- **Scores are conversations compressed.** The value was the argument about why Impact=2 not 3; a spreadsheet filled solo in an afternoon skipped the value and kept the theatre.
- **Re-run cadence:** quarterly, or when a strategy/market fact changes. Continuously re-scored backlogs are churn; never-re-scored ones are archaeology.

## Composite smells

- Everything scores medium → the scale isn't discriminating; force-rank the top 10 pairwise instead
- The ranking exactly matches the HiPPO's known preferences → the inputs were reverse-engineered; re-score blind
- Items keep being "re-scoped" until they rank → the framework is being gamed via the effort column; freeze definitions before scoring
