ICE Scoring
Prioritize with 3 criteria: Impact x Confidence x Ease. Sean Ellis' framework for fast, transparent and defensible decisions.
Description
ICE Scoring is a prioritization framework created by Sean Ellis that evaluates each idea on three criteria rated from 1 to 10 (Impact, Confidence, and Ease), then multiplies them to produce a ranking score. Its purpose: transform a chaotic backlog into an ordered priority list in under two hours, without political negotiation or endless debate. The problem is universal: your backlog overflows with customer requests, feature ideas, and technical debt, and everyone on the team defends their priorities with their own arguments. It is like a restaurant where every dish arrives at the pass at the same time, with no order of service. Sean Ellis, who coined the term "growth hacking" and former Head of Growth at Dropbox, developed ICE Scoring for his experimentation teams at GrowthHackers.com before formalizing it in his book Hacking Growth (2017, over 750,000 copies sold). The principle is radically simple: for each item, you rate the expected Impact (what gain if it works?), the Confidence (how certain are you of that impact?), and the Ease (how easy is it to execute?). The ICE score = Impact x Confidence x Ease produces an objective ranking where high-impact, high-certainty, low-effort items naturally rise to the top. The real strength of the framework lies in the Confidence criterion, which is often overlooked. Without it, you fall into the classic trap: overestimating the impact of your favorite ideas. Itamar Gilad has enriched this criterion with his Confidence Meter, which scales levels of evidence (opinion, secondary data, user test, A/B test). Compared to Intercom's RICE, which adds a Reach criterion, ICE Scoring favors decision speed over analytical completeness. That is a deliberate choice: a fast, revisable decision is better than a perfect analysis that takes three weeks.
Objectives
- Prioritize features
- Improve team collaboration
- Ensure strategic alignment
Used by
- -Dropbox (Sean Ellis developed the foundations of ICE Scoring there as Head of Growth before formalizing the method)
- -GrowthHackers.com (Sean Ellis's platform where ICE is the standard for experiment prioritization)
- -Intercom (used ICE Scoring before developing RICE, which adds the Reach criterion)
Advantages
- Speed of execution. An ICE workshop takes 1-2 hours maximum. You leave with an ordered roadmap.
- Mathematical transparency. The ICE score is reproducible and defensible. No more "why not my idea?", the score speaks.
- Makes saying "no" easy. Declining an idea becomes simple: "ICE score too low, we revisit next cycle."
- Adaptable to all contexts. Works as well for prioritizing product features as growth experiments or technical debt.
Limitations
- Subjective scoring despite the numbers. Scores remain human estimates. If the team is biased, the score will be too.
- Does not capture dependencies. An easy feature (high Ease) may depend on a complex one. ICE Scoring does not model these links.
- Ease bias. Quick wins with high ease dominate the ranking if you do not reserve space for ambitious bets.
- Confidence often underestimated. Teams score Impact and Ease carefully but rush through Confidence. Insist on it.
How to apply ICE Scoring
- 1
List all candidates to score
Empty your backlog: features, bugs, optimizations, explorations, growth experiments. Aim for 10-30 items maximum. Beyond that, segment by theme or by sprint. Output: a unified list visible to the entire team (Google Sheet, Notion, Linear).
- 2
Calibrate the scoring scale as a team
Before scoring, align everyone on what each rating means. Example: Impact 1 = marginal improvement, Impact 5 = measurable gain on a key metric, Impact 10 = game changer for the business. Do the same for Confidence (1 = pure intuition, 10 = validated by an A/B test) and Ease (1 = 3 months of dev, 10 = deployable in one day). Output: a shared reference grid to avoid interpretation gaps.
- 3
Score as a team, not solo
Gather PM, Tech Lead, Designer, and 1-2 business stakeholders. For each item, give a rating on each axis. If the gaps are too large (one dev says Ease = 2, another says 8), debate for 2 minutes then converge. The debate is the value: it reveals blind spots. Output: table filled with the 3 scores per item.
- 4
Emphasize the Confidence criterion
This is the most important and most neglected criterion. Ask the question: "What is our Impact estimate based on?" If the answer is "we think that...", Confidence is low (2-3). If it is "the data shows that...", it is high (7-8). An item with Impact 10 but Confidence 2 should not outrank an item with Impact 6 and Confidence 9. Output: honest and documented Confidence scores.
- 5
Calculate the ICE score
Formula: ICE Score = Impact x Confidence x Ease. Example: Impact 8 x Confidence 7 x Ease 6 = 336. Sort your list by descending score. The items at the top are your objective priorities. Output: list sorted by ICE score.
- 6
Validate the top 5-7 items
Look at the 5-7 highest-scoring items. Verify that they are consistent with your product strategy and that there are no blocking dependencies. If a high-scoring item surprises you, now is the time to challenge it: did you score correctly, or is your intuition biased? Output: top items validated strategically.
- 7
Reserve 20% for "big swings"
Classic ICE Scoring trap: easy items (high Ease) and safe items (high Confidence) dominate the top of the ranking, at the expense of high-impact but uncertain bets. Reserve 20% of your capacity for high-Impact but low-Confidence items. That is where breakthrough innovations hide. Output: explicit allocation between "quick wins" and "big bets."
- 8
Rescore regularly (every month)
Priorities evolve: new user data, customer feedback, technical constraints. Rescoring takes only 30 minutes once the system is in place. A frozen ICE score is a dead score. Output: living backlog updated monthly.