ICE-based Idea Prioritization (Evidence-Guided)

Goal

Ingest an idea description and current state, score it on Impact, Confidence, and Ease, and compute the ICE Score = Impact × Confidence × Ease to propose an execution priority. All evaluations must follow explicit criteria and rely only on stated evidence—no inference or guesswork.

---

When to Use

• To quickly order an idea backlog and select items for exploration and experiments
• When you have at least some explicit evidence (interviews/data/tests) and a rough team effort estimate
• Before drafting experiment plans for a leaf opportunity in an Opportunity-Solution Tree

---

Input

• Idea title and description: goal, target metric, scope, and working hypothesis
• Idea analysis and current state: data/user/market/test evidence, execution hypothesis, risks, and effort estimate
• Optional:
  • Target metric and expected change rate (%) or range
  • Estimated effort (person-weeks)

---

Output

• Format: Markdown (.md)
• Location: initiatives/[initiative]/solutions/
• Filename: ice-[YYYY-MM-DD]-[slugified-idea-title].md

---

Scoring Model

1) Impact Mapping

Target Metric Change (%)	Impact
> 50%	10
35 - 49.9%	9
25 - 34.9%	8
18 - 24.9%	7
12 - 17.9%	6
7 - 11.9%	5
4 - 6.9%	4
2 - 3.9%	3
0.5 - 1.9%	2
0.1 - 0.4%	1
≤ 0%	0

• Missing data handling: If no explicit percentage is provided, first ask the user for an estimate. If the user cannot provide one, apply default +1.5% improvement (Impact 2) and add a warning in the output: ⚠️ DEFAULT VALUE: Impact uses assumed +1.5% improvement due to missing data.

2) Ease Mapping (Estimated Effort: person-weeks)

Duration	Ease
< 1 week	10
1–2 weeks	9
3–4 weeks	8
5–6 weeks	7
6–7 weeks	6
8–9 weeks	5
10–12 weeks	4
13–16 weeks	3
17–25 weeks	2
≥ 26 weeks	1

3) Evidence Types (Count only evidence directly tied to Impact)

Evidence Type	Description
Test Results	A/B tests, longitudinal user studies, beta experiments, large MVPs with quantitative validation
User-based Evidence	Product usage data, 20+ user interviews, usability studies, MVP results/feedback
Market Data	Surveys, smoke tests, "table stakes" in the competitive set
Empirical Evidence	Few data points, sales requests, 1–3 interested customers, one competitor has the feature
Estimates & Plans	Internal model-based estimates, feasibility review with Eng/Design, schedule/business model analysis
Opinions of Others	Executives/colleagues/experts/investors opinions
Directional Fit	Alignment with company vision/strategy, tech/market trends, external research, macro trends
Self-belief	Personal intuition/gut feel/experience

• Caution: Use only explicitly stated evidence in the input. No inference.
• Statements like "intuitively", "personally I think", "my gut says" → classify as Self-belief.
• Without explicit quantitative backing, do not accept as Market Data or Estimates & Plans.

4) Confidence Calculation

• Principle: Include only evidence that directly supports Impact.
• Per-type contribution = MIN(Weight × count, Max)
• Group caps (sum upper bounds):
  • Self-belief + Directional Fit ≤ 0.1
  • Opinions of Others + Estimates & Plans ≤ 0.5
  • Market Data + User-based Evidence ≤ 3.0

Evidence Type	Weight	Max
Self-belief	0.01	0.1
Directional Fit	0.05	0.1
Opinions of Others	0.10	0.5
Estimates & Plans	0.30	0.5
Empirical Evidence	0.50	1.0
Market Data	1.0	3.0
User-based Evidence	2.0	3.0
Test Results	3.0	5.0

• Keyword hints (examples): "test/experiment/AB", "user request/behavioral data", "market/competitor/table stakes", "estimate/modeling", "intuition/gut/personally".

5) ICE Score and Priority Interpretation

• Formula: ICE = Impact × Confidence × Ease
• Interpretation:
  • ≥ 250: Consider immediate execution (high expected ROI)
  • 150–249: Promising; recommend additional precision testing
  • 100–149: Proceed with mitigations or phase-two testing
  • < 100: On hold or needs strengthening

---

Process

1. Input validation

• Verify target metric, expected change (%), evidence text, and estimated effort (person-weeks)
• If missing, ask clarifying questions about metric/change, effort, and evidence type/source

2. Impact scoring

• Map % change to table; if missing, apply default Impact 2

3. Ease scoring

• Map person-weeks to table; if uncertain, use conservative lower ease

4. Evidence extraction and classification

• Count only Impact-related evidence from the input
• Tally per type and apply group caps

5. Confidence calculation

• Sum per-type contributions → apply group caps → final Confidence (0–10)

6. ICE computation and bucket

• Compute ICE = I × C × E; assign interpretation bucket

7. Report generation

• Include score table, calculation rationale, cap applications, risks/assumptions, and recommended next steps

---

Output Format

# ICE Evaluation — [Idea Title]

## Overview
- **Idea:** [Title]
- **One-line Summary:** [Brief description]
- **Target Metric:** [Metric name]
- **Assumptions/Scope:** [Key assumptions]

## Score Summary
- **Impact:** [I] (basis: [expected % change or default rule])
- **Ease:** [E] (basis: [person-weeks])
- **Confidence:** [C]
  - Details:
    - Self-belief: 0.01 × [n] → [x] (max 0.1, Group A ≤ 0.1)
    - Directional Fit: 0.05 × [n] → [x]
    - Opinions of Others: 0.10 × [n] → [x] (Group B ≤ 0.5)
    - Estimates & Plans: 0.30 × [n] → [x]
    - Empirical Evidence: 0.50 × [n] → [x] (max 1.0)
    - Market Data: 1.0 × [n] → [x] (Group C ≤ 3.0)
    - User-based Evidence: 2.0 × [n] → [x]
    - Test Results: 3.0 × [n] → [x] (max 5.0)
  - Group caps applied:
    - Group A (Self-belief + Directional Fit): [sum] → [capped]
    - Group B (Opinions + Estimates): [sum] → [capped]
    - Group C (Market + User): [sum] → [capped]
  - **Final Confidence:** [C]

## ICE Calculation
- ICE = [I] × [C] × [E] = **[Score]**
- **Priority Guidance:** [Bucket label]

## Input Summary
- **Expected Metric Change:** [value/none → default 1.5% applied]
- **Estimated Effort (person-weeks):** [value/uncertain]
- **Evidence Excerpts:** 
  - [Excerpt 1 — classified as: user/market/test/...]
  - [Excerpt 2 — classified as: ...]

## Risks / Assumptions
- [Key risk]
- [Key uncertainty]
- [Critical assumption]

## Recommended Next Steps
- [Tests/data collection/research/prototype]
- Confidence Improvement Plan: [Which evidence to strengthen]

## Notes
- ICE is a fast comparison/sorting tool; final decisions must also consider strategy, market, and resources.

---

Example

• Idea: "AI-based revenue anomaly detection dashboard"
• Expected change: 26% → Impact 8
• Effort: 5 weeks → Ease 7
• Confidence input:
  • User-based Evidence (3) → 2.0 × 3 = 6.0 → Group C cap (≤ 3.0) → 3.0
  • Estimates & Plans (2) → 0.3 × 2 = 0.6 → Group B cap (≤ 0.5) → 0.5
  • Total: 3.0 + 0.5 = 3.5 → Final C = 3.5
• ICE = 8 × 3.5 × 7 = 196 → "Promising; recommend additional precision testing"

---

Customization (Team Tuning)

• Adjust Impact bands to your target metric sensitivity
• Adjust Ease bands to team speed/role mix
• Extend evidence keywords to your domain language, while preserving the "explicit evidence only" rule
• Recalibrate bucket thresholds per quarterly capacity/roadmap density

---

Guardrails

• Do not invent/infer evidence or over-credit weak signals
• Exclude evidence not directly tied to Impact from Confidence
• When uncertain, apply conservative caps and document in "Risks/Assumptions"
• Prevent duplicate counting of the same source/content

---

Error Handling

• Missing % change/effort: apply default rules (Impact 2) or ask clarifying questions
• Insufficient evidence: request the needed evidence types with examples
• Conflicting info: note conflicts and dependent assumptions; use conservative scoring

---

Interaction Model

1. Collect & validate inputs → 2) Score I/E → 3) Classify/count evidence → 4) Compute Confidence (with caps) → 5) Compute ICE & assign bucket → 6) Generate report → 7) Resolve gaps/uncertainties and update

---