ICE Framework: Fast Feature Prioritization

Use ICE to compare a constrained backlog

Product backlogs often contain more ideas than a team can build. ICE offers a quick way to compare them using impact, confidence, and ease.

ICE scoring

ICE uses three questions and one calculation.

The Formula:

ICE Score = Impact × Confidence × Ease

1. Impact: How much will this help?

Rate from 1 to 10. A higher score indicates a larger potential effect on your main goal.

Examples:

• 10: A very large expected effect on the main metric
• 7: A substantial expected improvement
• 5: Nice improvement users will notice
• 3: Small enhancement
• 1: Barely noticeable

What to ask: "If this works as intended, how much will it move our main metric?"

2. Confidence: What evidence supports it?

Rate from 1 to 10. A higher score indicates stronger evidence that the feature will have its expected effect.

Quick Examples:

• 10: An experiment or data strongly supports the expected effect
• 7: Multiple customers asked for it
• 5: Makes logical sense, no data yet
• 3: Interesting idea, might work
• 1: Complete guess

What to ask: "What evidence supports this estimate?"

3. Ease: How easy is it to implement?

Rate from 1 to 10. A higher score indicates that the work is easier or faster to complete.

Quick Examples:

• 10: Very small change
• 7: Simple feature
• 5: Standard development task
• 3: Complex feature
• 1: Major project

What to ask: "What engineering and design effort does this require?"

Try It Now

Score a set of features:

Example exercise

Score these for a B2B SaaS:

Feature A: Dark Mode

• Impact: 4 (nice to have)
• Confidence: 8 (users requested it)
• Ease: 7 (CSS changes)
• Score: 224

Feature B: 50% Faster Dashboard

• Impact: 9 (affects everyone daily)
• Confidence: 7 (bottlenecks identified)
• Ease: 5 (backend refactoring)
• Score: 315

Feature C: AI Recommendations

• Impact: 8 (could differentiate)
• Confidence: 3 (untested assumption)
• Ease: 2 (complex, risky)
• Score: 48

The performance improvement receives the highest score in this example.

Real-World Examples

Airbnb: Professional Photos

The situation: Listings weren't getting bookings. Growth was flat.

The idea: Hire photographers to take professional listing photos.

ICE Scoring:

• Impact: 8 (better photos should mean more bookings)
• Confidence: 6 (made sense but no data yet)
• Ease: 7 (just hire local photographers)
• Score: 336

Why these scores: The example treats photos as high impact, assigns medium confidence because the idea was untested, and treats the operational work as relatively easy.

Result: The source reports that bookings increased after the experiment¹.

Buffer: Price Comparison Table

The situation: Users kept asking how Buffer compared to competitors.

The idea: Add a comparison table to the pricing page.

ICE Scoring:

• Impact: 7 (removes purchase friction)
• Confidence: 8 (users literally asked for this)
• Ease: 9 (one day of work)
• Score: 504

Why these scores: Direct user requests support the confidence estimate, and the team considered the work relatively easy.

Result: Buffer reported that the table reduced pricing questions and clarified the upgrade path.

Superhuman: Keyboard Shortcuts

The situation: Building "the fastest email experience ever."

The idea: Add keyboard shortcuts for every action.

ICE Scoring:

• Impact: 9 (core to their value proposition)
• Confidence: 7 (power users in beta loved it)
• Ease: 4 (lots of engineering work)
• Score: 252

Why these scores: Core feature for their brand. Good user feedback. Complex to build properly.

Result: The score was lower, but Superhuman built it anyway because it was core to the brand, and it became their signature feature.

ICE vs RICE: When to Use Each

ICE is generally faster to apply. RICE adds a Reach estimate. Choose based on the decision and available data.

Use ICE When:

• Small team
• Need a quick ranking
• Building your MVP
• Limited user data
• Testing ideas quickly

Use RICE When:

• Larger team
• Planning quarterly roadmaps
• Have detailed user segments
• Need stakeholder buy-in
• Lots of user data available
• Making big investment decisions

How to Transition:

1. Start with ICE when you're small and fast
2. Add "Reach" when you understand user segments
3. Consider RICE when you can estimate Reach reliably
4. Keep using ICE for quick experiments

Use ICE when a consistent, fast ranking is more useful than a detailed Reach estimate. Add RICE when Reach data would materially change the decision.

Common Pitfalls

1. Everything Scores 7

Problem: Middle bias. Everything seems equal. Fix: Force rank: 1, 3, 5, 7, 10. No repeats in one session.

2. Ease Becomes Effort

Problem: Confusing ease (10=easy) with effort (10=hard). Fix: Write the scale direction beside the score. In ICE, 10 means easier.

3. Impact Without Goals

Problem: Impact becomes vague. Fix: Define your metric first. "Impact on 7-day retention."

4. Confidence Inflation

Problem: Optimism bias. Everything gets high confidence. Fix: Set score definitions in advance and use evidence to support higher confidence scores.

5. Stale Scores

Problem: Score once, never update. Fix: Re-score priorities when material evidence or circumstances change.

AI Prompts for Better ICE Scoring

Team Score Calibration

Help my team align on ICE scoring standards.
Our product: [describe briefly]
Main goal: [retention/growth/revenue]

Create a scoring rubric with specific examples for:
- Impact: What does 1, 5, and 10 look like for our goal?
- Confidence: What evidence qualifies for each score?
- Ease: How many days of work for each score level?

Make examples specific to our product context.

Bias Detection and Prevention

Review these ICE scores for common biases:
[paste your scored features]

Check for:
1. Middle bias (everything scoring 5-7)
2. Optimism bias (confidence too high without data)
3. Ease confusion (mixing ease with effort)
4. Recency bias (new ideas scoring higher)

For each bias found:
- Flag the problematic scores
- Explain why it's biased
- Suggest corrected score with reasoning

Quick Feature Triage

Score these feature requests using ICE:
[paste feature list]

Context:
- Stage: [MVP/Growth/Scale]
- Team size: [number]
- Main metric: [what you optimize for]
- Sprint length: [1 or 2 weeks]

For each feature:
1. Assign Impact, Confidence, Ease (1-10)
2. Calculate ICE score
3. One-line reasoning for scores
4. Flag any that need more research

Sort by ICE score and highlight top 3 to build.

Past Score Analysis

Analyze our past ICE scoring accuracy:

Features we shipped last quarter:
[Feature name | ICE scores | Actual outcome]

Questions to answer:
1. Which component (I, C, or E) were we worst at estimating?
2. What patterns explain our misses?
3. How should we adjust our scoring going forward?
4. Specific calibration recommendations

Generate a "lessons learned" checklist for future scoring.

Stakeholder Decision Rationale

Generate a clear explanation for this prioritization decision:

Feature A: [description] - ICE Score: [X]
Feature B: [description] - ICE Score: [Y]
We're choosing: [A or B]

Create a 3-paragraph explanation covering:
1. Why the winner scored higher (break down I, C, E)
2. What evidence supports this decision
3. When we'll revisit the deprioritized feature

Keep language non-technical for executive audience.

How to Run a Team ICE Session

Use a short session when a quick ranking is sufficient:

Before the meeting

• Pick 10 features maximum
• Write one-line descriptions
• Share the 1-10 scale definitions

During the Meeting

Step 1: Silent scoring Have everyone score the features independently before discussion. This can reduce the influence of early opinions.

Step 2: Compare and discuss Show the scores and focus discussion on items with meaningful disagreement. Ask participants to explain their reasoning.

Step 3: Final scores Re-score after discussion if needed, then agree how the team will combine scores and rank the ideas.

Tips for Success

• Set a timer for each section
• Write scores down, don't just think them
• Decide how to resolve ties before the session

Scoring on Your Own

When you need to score quickly by yourself:

1. Impact - Initial estimate (20 seconds)
2. Confidence - What evidence do you have? (20 seconds)
3. Ease - Quick estimate or ask engineer (20 seconds)

Use a short, consistent time limit if speed is the purpose of the exercise.

ICE Variations

Weighted ICE (WICE)

Add 1.5x multiplier for OKR-aligned features. Use sparingly.

Category ICE

Separate lists for Features, Bugs, Tech Debt. Compare apples to apples.

Time-Boxed ICE

Ease becomes binary: "Can we ship this sprint?" (Yes=10, No=1).

Customer ICE

Customers score Impact. Confidence = number requesting it.

Key Takeaways

• Use ICE to compare ideas by impact, confidence, and ease.
• Define each score level before scoring as a team.
• Treat scores as a decision aid and revisit them when evidence changes.

Next Steps

Master rapid prioritization:

1. Score features with our ICE Calculator
2. Compare with RICE Scoring for complex decisions
3. Plot on Impact/Effort Matrix for visualization
4. Try Weighted Scoring for custom criteria

Sources