Understanding ICE Scores
Master the ICE scoring framework to prioritize experiments effectively and maximize your growth experimentation ROI.
What is ICE Scoring?
ICE scoring is a simple yet powerful prioritization framework that helps growth teams decide which experiments to run first. Each experiment is rated on three dimensions on a 1–10 scale, and the average determines your overall priority score.
- Impact - How significantly could this experiment move your key metrics?
- Confidence - How certain are you that this experiment will produce the expected result?
- Ease - How simple is this experiment to implement?
Impact: Measuring Potential Upside
| Score | Meaning |
|---|---|
| 9-10 | Game-changing, could significantly move key metrics |
| 7-8 | Strong potential to improve metrics meaningfully |
| 5-6 | Moderate improvement expected |
| 3-4 | Small incremental gain |
| 1-2 | Minimal impact, mostly learning |
Confidence: Assessing Likelihood
| Score | Meaning |
|---|---|
| 9-10 | Strong evidence, data-backed hypothesis |
| 7-8 | Good evidence from similar tests or industry data |
| 5-6 | Reasonable assumption based on experience |
| 3-4 | Gut feeling, limited evidence |
| 1-2 | Wild guess, pure experiment |
Ease: Evaluating Implementation
| Score | Meaning |
|---|---|
| 9-10 | Can do in a few hours, no dependencies |
| 7-8 | A day or two of work |
| 5-6 | About a week |
| 3-4 | Multiple weeks, some dependencies |
| 1-2 | Large effort, many dependencies |
The Smart ICE Editor
Instead of manually guessing scores, the Smart ICE Editor provides AI calibration:
- Open any experiment and navigate to the ICE score section
- Click "Get AI Suggestions"
- Review per-dimension reasoning and a priority tier recommendation (High / Medium / Low)
- Click "Apply All" to accept the AI scores, or apply individual scores per dimension
- Adjust based on your domain knowledge
💡 Tip: The Smart ICE Editor factors in past experiment learnings. If similar experiments succeeded or failed before, the AI adjusts confidence scores accordingly and shows the historical influence.
What the AI Considers
- Your business context (industry, stage, resources)
- Historical experiment data and success patterns
- Industry benchmarks for similar tactics
- Complexity signals from your hypothesis and experiment map
- Learning influence - past experiments that inform confidence calibration
Learning Calibration
When the AI detects relevant past learnings, each score dimension may show a learning influence badge indicating:
- The modifier applied (e.g., +2 or -1)
- The source learning that informed the adjustment
- Success rate of similar past experiments
