rice-scoring-patterns

name	rice-scoring-patterns
description	RICE scoring framework, effort estimation, and impact modeling for feature prioritization. Use when scoring a feature backlog, estimating effort, or modeling expected outcomes before committing to a development cycle.

RICE Scoring Patterns

RICE is a structured framework for scoring features against each other so that prioritization decisions are explicit and defensible. It forces you to estimate four dimensions — Reach, Impact, Confidence, Effort — and produces a comparable score for each feature.

When to Use This Skill

• Scoring a backlog of features before a planning cycle
• Comparing investment options with different reach/impact profiles
• Communicating prioritization rationale to stakeholders
• Identifying quick wins (high RICE score, low effort)
• Flagging overconfident estimates that need validation

The RICE Formula

RICE Score = (Reach × Impact × Confidence) ÷ Effort

Higher score = higher priority.

Reach

Definition: How many users will this affect per time period (typically per quarter)?

• Count users who will encounter the feature, not total users
• Be specific: "users who complete checkout" not "all users"
• Source from analytics data where possible; flag when estimated

Examples:
- "All logged-in users" → 50,000 users/quarter
- "Users who reach the checkout screen" → 12,000 users/quarter
- "Sellers with >100 listings" → 800 users/quarter

Impact

Definition: How much does this affect each user who encounters it? Use a fixed scale:

Score	Definition
3	Massive impact — transforms the experience or drives a step-change in the key metric
2	High impact — meaningfully improves a key metric
1	Medium impact — noticeable improvement
0.5	Low impact — minor improvement
0.25	Minimal impact — very small effect

Important: Impact is relative to the outcome being optimized (conversion, retention, revenue, engagement), not just user satisfaction.

Confidence

Definition: How confident are you in the Reach and Impact estimates? Express as a percentage:

Score	Definition
100%	Strong evidence — user research + data + prior experiments
80%	Good evidence — user research or data, but not both
50%	Moderate evidence — directional signals but no hard data
20%	Low evidence — hypothesis based on intuition

Rule: If confidence is below 50%, consider a validation experiment before the full build.

Effort

Definition: Total person-months required across all roles (engineering, design, PM, QA).

• 1 person working for 1 month = 1
• 2 engineers + 0.5 designer for 3 weeks = approximately 1.4
• Use calibrated estimates against historical velocity where possible

Note: Effort uses person-months, not story points or arbitrary units. This makes RICE scores comparable across teams with different sprint structures.

Worked Example

Feature	Reach	Impact	Confidence	Effort	RICE Score
Saved payment methods	12,000	3	80%	2	14,400
Guest checkout	8,000	2	100%	1	16,000
Order status push notifications	50,000	0.5	80%	0.5	40,000
Bulk listing upload for sellers	800	3	50%	3	400

Reading: Guest checkout and push notifications score much higher than they might feel intuitively — guest checkout has a high confidence that boosts it; push notifications win on Reach despite low Impact per user.

Effort Estimation Patterns

T-Shirt Sizing with Person-Month Conversion

T-Shirt	Person-Months	Characteristics
XS	0.25	Config change, copy update, minor UI tweak
S	0.5–1	Single-surface change, existing patterns, no new APIs
M	1–2	Multi-surface, some new backend logic, standard integration
L	3–5	New subsystem, significant backend work, multiple integrations
XL	6+	Platform-level change, new data model, cross-team dependencies

Dependency-Adjusted Effort

When Feature B requires Feature A first, the total effort for B includes the dependency overhead:

• Add 20–30% for integration work between A and B
• Add sequencing cost: B can't start until A is done, which may delay B by a full sprint

Risk-Adjusted Estimates

Always provide three estimates:

Optimistic: Everything goes smoothly, no surprises
Expected:   Normal velocity, 1-2 minor surprises
Pessimistic: Key dependency is harder than expected, one rework cycle

Expected estimate = (Optimistic + 4×Most Likely + Pessimistic) ÷ 6

Common RICE Mistakes

Mistake	Effect	Fix
Using total users for Reach	Inflates scores for features with narrow real reach	Scope Reach to users who will actually encounter the feature
Overconfidence (everyone scores 80%+)	Prioritization feels certain when it isn't	Force calibration: what evidence supports each confidence score?
Ignoring second-order effects	Undervalues prerequisite features	Add "prerequisite multiplier" — if A enables B and C, A's effective value includes B and C
Effort in story points	Scores aren't comparable across teams	Convert to person-months
One-time scoring	Backlog gets stale as evidence accumulates	Re-score after each validation experiment

Outcome Modeling

For high-stakes items (large effort, strategic importance), complement RICE with a simple outcome model:

Feature: [Name]
Target metric: [e.g., checkout conversion rate]
Current baseline: [e.g., 62%]

Scenario modeling:
  Optimistic: +5pp conversion → +$840K ARR at current volume
  Base case:  +2.5pp conversion → +$420K ARR
  Pessimistic: +1pp conversion → +$168K ARR

Expected value: 0.2 × $840K + 0.6 × $420K + 0.2 × $168K = $453K ARR
Engineering cost: 2 person-months × $15K/month blended = $30K
Expected ROI: 15:1

Best Practices

1. Score collaboratively — PM + engineering + design should each estimate effort; take the average or discuss divergence
2. Calibrate against history — compare previous estimates to actuals and adjust systematically
3. Flag low-confidence estimates — a high RICE score with 20% confidence is a hypothesis, not a priority
4. Re-score after experiments — if a validation experiment changes your Impact or Confidence estimate, the ranking changes
5. RICE is an input, not the output — the stack rank is a recommendation; human judgment still overrides for strategic, dependency, or constraint reasons