02 · Prioritization Frameworks Deep Dive¶
In Level 1 you learned RICE and MoSCoW, which are enough to sort a backlog of comparable items. They start failing on the decisions that actually cost you a quarter: a feature everyone asks for that adds nothing to satisfaction, a boring reliability fix that outranks the exciting bet, two items with the same score and different risk profiles. This module adds four tools for those cases — weighted scoring for multi-criteria tradeoffs, Kano for what a feature does to satisfaction, opportunity scoring for which problem to attack at all, and WSJF / cost of delay for sequencing when timing matters more than size.
The meta-skill is knowing that a framework is a conversation structure, not a calculator. Its job is to make disagreement specific. When two frameworks rank the same list differently, that gap is the decision — and this module ends by showing exactly that.
Weighted scoring¶
RICE hard-codes one opinion about what matters. Weighted scoring makes your criteria and their weights explicit, so you argue about the weights once, in the open, instead of re-litigating every item.
Steps: pick 4–6 criteria → assign weights summing to 1.0 → score each item 1–5 per criterion → total = Σ(score × weight).
| Criterion | Weight | 1 means | 5 means |
|---|---|---|---|
| Outcome impact | 0.35 | No plausible link to the quarter's outcome | Directly moves the outcome metric |
| Reach | 0.20 | Under 10% of active accounts | Nearly every active account |
| Confidence | 0.15 | One anecdote | Interviews + analytics + support agree |
| Effort (inverted) | 0.20 | Multi-quarter | Under a week |
| Strategic fit | 0.10 | Off-strategy; we'd have to explain it | Extends our stated wedge |
Two rules keep this honest. Weights are set before items are scored — otherwise people reverse-engineer weights to favour a pet feature. And effort is inverted so every criterion points the same way (higher = better); mixing directions is the most common arithmetic error in homemade scoring models.
The Kano model¶
Kano classifies a feature by what it does to satisfaction — which is not proportional to how loudly customers ask for it.
| Category | If present | If absent | Investment rule |
|---|---|---|---|
| Must-be | No credit | Deal-breaker | Meet the bar, stop. Polish buys nothing |
| Performance | Better, roughly linearly | Proportionally worse | Invest where you compete — this is the scoreboard |
| Attractive | Delight, disproportionate | Nobody notices | A few, deliberately. This is differentiation |
| Indifferent | Nothing | Nothing | Cut. This is where roadmaps quietly die |
| Reverse | Actively worse | Better without it | Remove, or make it optional |
You classify by asking each user a question pair about the same feature:
| Question | Answer options |
|---|---|
| Functional: "If ListUp showed a price-change preview before publishing, how would you feel?" | I like it / I expect it / I'm neutral / I can tolerate it / I dislike it |
| Dysfunctional: "If ListUp did not show a price-change preview, how would you feel?" | (the same five) |
The pair maps to a category: like it + dislike its absence = Performance; like it + neutral about its absence = Attractive; expect it + dislike its absence = Must-be; neutral + neutral = Indifferent.
The decay rule matters more than any single classification: attractive features become performance features, then must-be features, usually within a few years. Re-run Kano on your core set roughly annually, or you will keep paying delight prices for table stakes.
Opportunity scoring¶
Kano and weighted scoring rank solutions. Opportunity scoring tells you which problem deserves a solution. Survey users on two 1–10 scales per job step — how important it is, and how satisfied they are with today's options:
Opportunity = Importance + max(Importance − Satisfaction, 0)
Roughly: above 10 is underserved, below 7 is served — and often over-served, which is a signal to cut investment, not add.
| Job step | Importance | Satisfaction | Opportunity | Read |
|---|---|---|---|---|
| Trust that a price edit reached every channel | 9.1 | 4.2 | 14.0 | Badly underserved — fund it |
| Connect a second marketplace quickly | 8.2 | 5.1 | 11.3 | Underserved |
| Import without losing my own notes | 7.4 | 5.8 | 9.0 | Borderline |
| Bulk-edit listing titles | 6.0 | 5.9 | 6.1 | Served — stop investing |
Cost of delay and WSJF¶
Everything above ranks by value and size. None of it asks what it costs to ship later. Cost of delay (CoD) is value lost per unit of delay, and WSJF = CoD ÷ job size sequences work so total delay cost is minimized.
| CoD component | Question | Scale |
|---|---|---|
| User/business value | What do we lose per month without it? | 1–10 |
| Time criticality | Does the value decay? Is there a fixed date? | 1–10 |
| Risk reduction / opportunity enablement | Does it unblock later work or kill a big unknown? | 1–10 |
Score in relative terms (1, 2, 3, 5, 8, 13). WSJF only needs the ranking to be right, not the units.
| Item | Value | Time crit. | Risk/enable | CoD | Job size | WSJF |
|---|---|---|---|---|---|---|
| Sync failure alerts | 5 | 8 | 2 | 15 | 2 | 7.5 |
| Pre-publish diff preview | 8 | 5 | 8 | 21 | 5 | 4.2 |
| CSV column mapper | 3 | 1 | 2 | 6 | 3 | 2.0 |
| OAuth marketplace connect | 5 | 3 | 3 | 11 | 8 | 1.4 |
Choosing a framework¶
| Situation | Use | Why |
|---|---|---|
| Comparable items, one team, one quarter | RICE | Fast, one number, everyone knows it |
| Stakeholders disagree about what matters | Weighted scoring | Forces the weights argument into the open, once |
| "Customers asked for it" is driving the roadmap | Kano | Separates asked-for from satisfaction-moving |
| You don't know which problem to attack | Opportunity scoring | Ranks outcomes, not features |
| Fixed dates, decaying value, dependencies | WSJF | The only one that prices lateness |
| Two options, high stakes, thin data | None — run an assumption test | A score cannot manufacture evidence |
Worked example — ListUp¶
Five candidate solutions from the opportunity solution tree in Module 1, run through three frameworks over the same quarter.
RICE — Reach = accounts touched per quarter, Impact 0.25–3, Effort in person-months, score = R × I × C ÷ E:
| Item | R | I | C | E | Score |
|---|---|---|---|---|---|
| Sync failure alerts | 2,400 | 0.5 | 100% | 0.5 | 2,400 |
| Pre-publish diff preview | 1,800 | 2.0 | 80% | 2.0 | 1,440 |
| Sync status badge | 2,400 | 1.0 | 80% | 1.5 | 1,280 |
| CSV column mapper | 900 | 1.0 | 100% | 1.0 | 900 |
| OAuth marketplace connect | 1,500 | 2.0 | 50% | 3.0 | 500 |
Weighted scoring with the weights above (1–5, effort already inverted):
| Item | Impact ×.35 | Reach ×.20 | Conf ×.15 | Effort ×.20 | Fit ×.10 | Total |
|---|---|---|---|---|---|---|
| Pre-publish diff preview | 5 | 4 | 4 | 3 | 5 | 4.25 |
| Sync status badge | 3 | 5 | 4 | 4 | 3 | 3.75 |
| Sync failure alerts | 2 | 5 | 5 | 5 | 3 | 3.75 |
| OAuth marketplace connect | 4 | 3 | 3 | 2 | 4 | 3.25 |
| CSV column mapper | 2 | 2 | 5 | 5 | 2 | 3.05 |
The frameworks disagree, and that is the finding. RICE and WSJF both put sync failure alerts first; weighted scoring puts diff preview first. The gap has one cause: RICE multiplies broad reach by cheap effort, rewarding a small change that touches everyone, while the weighted model puts 45% of its weight on outcome impact and strategic fit — where alerts score 2 and 3.
Kano settles it. From a 40-seller survey:
| Feature | Modal functional / dysfunctional answer | Category |
|---|---|---|
| Sync failure alerts | "I expect it" / "I dislike its absence" | Must-be |
| Pre-publish diff preview | "I like it" / "I'm neutral" | Attractive |
| CSV column mapper | "I'm neutral" / "I'm neutral" | Indifferent |
Must-be features never raise satisfaction — but their absence caps it. So the decision isn't which one, it's sequence: ship alerts first, because it is a half-person-month must-be that is currently missing and you cannot buy delight below the floor; then spend the rest of the quarter on the diff preview, the only attractive-category item and the only one scoring 5 on the quarter's outcome. The CSV mapper — the highest-confidence item in the RICE table — gets cut, because Kano says nobody will notice it. Confidence that a feature will land is not evidence that it matters.
How It Actually Works¶
Weighted scoring frameworks (RICE, ICE, etc.) don't just rank items — they change which items win by altering the functional form of the tradeoff. RICE = (Reach × Impact × Confidence) / Effort is a multiplicative model, which means it mathematically punishes items that are weak on any single factor far more harshly than an additive model would — a huge-impact idea with near-zero reach still scores low, correctly reflecting that impact without reach is not organizational impact. This is also why RICE is sensitive to Confidence as a discount factor: multiplying by a 20%-80% confidence score explicitly shrinks the expected value of unvalidated bets, which mathematically encodes the same discounting a rational expected-value calculation would apply to a risky asset, and is the reason RICE tends to systematically favor validated, incremental work over speculative moonshots unless Impact is scored high enough to compensate — a known distortion worth watching for. The Kano model works by exploiting a real, measured psychological asymmetry: satisfaction from "delighters" is unbounded and grows with investment, while dissatisfaction from missing "must-haves" is a step function (any amount below threshold causes disproportionate anger) — this is loss aversion showing up in feature perception, and it's why teams that only track "average satisfaction" miss that under-investing in a must-have can tank retention even while overall sentiment scores look fine. Cost of Delay / WSJF works because it makes the time value of a decision explicit — economically, delaying a high cost-of-delay item is not neutral, it's actively destroying value every week (like unpaid interest accruing), while WSJF's job-size denominator ensures short jobs with high urgency get pulled forward even over larger, more "important-sounding" ones, correctly modeling that clearing a fast queue often beats context-switching onto one big item.
Exercise¶
Take your own backlog (or ListUp's five candidates) and produce:
- A weighted scoring model: 4–6 criteria with weights summing to 1.0, each with written definitions of what 1 and 5 mean. Get one stakeholder to agree to the weights before anything is scored, and note where they pushed back.
- Rankings from two frameworks over the same 5+ items (weighted scoring plus RICE or WSJF). Wherever the rankings differ, write one sentence naming the specific input that caused the difference.
- A Kano question pair for your top three items, run past at least five users. Classify each and state what the classification changes.
- An opportunity score table for 4+ steps of your users' main job. Name one over-served step you should stop investing in, and where that capacity goes instead.
- A one-paragraph decision memo: what you're building, what you're cutting, and — most importantly — which framework you overruled and why.