ICE Scoring is a simplified alternative to RICE, optimised for speed. By using Ease instead of Effort (inverted), it produces a quick composite score that growth teams use to triage experiments without detailed estimation.
Which experiments or ideas should we run next, based on impact, confidence, and ease?
impact * confidence * easeimpact * confidence * easeimpact * confidence * ease| Item | Impact | Confidence | Ease | Score▼ |
|---|---|---|---|---|
Checkout simplifyfeature | 8 | 8 | 7 | 448 |
Push notificationsfeature | 6 | 9 | 8 | 432 |
Referral programmefeature | 8 | 7 | 6 | 336 |
Homepage redesignexperiment | 7 | 8 | 5 | 280 |
AI recommendationssolution | 9 | 5 | 3 | 135 |
impact * confidence * easeICE Scoring ranks ideas, experimentsExperimentValidationA test designed to validate a hypothesisView reference →, or initiativesInitiativeStrategyA large coordinated effort to achieve a strategic goalView reference → by three factors: Impact (how much will this move the metricMetricStrategyA unified metric that measures progress, health, or behaviour across the productView reference → if it works), Confidence (how sure are we it will work), and Ease (how cheap is it to do). Each factor is scored 1-10; the composite is multiplied or averaged. Sean Ellis popularised ICE in growth-team practice as a faster, less formal alternative to RICE.
Reach for ICE when the team has more candidate experiments than capacity and needsNeedUserA user need, pain, desire, or constraintView reference → a quick way to triage. ICE is deliberately rough: the goal is to surface the obviously-good and obviously-bad ideas in twenty minutes, not to engineer a perfect ranking. When the cost of estimation exceeds the cost of just running the experiment, ICE is the right shape.
ICE is a view definition that ranks UPG entities, typically experimentsolutionopportunity
Choose the entity type you're scoring (usually a list of candidate experiments). For each candidate, score Impact 1-10 (the size of the win if it works), Confidence 1-10 (the strength of the evidenceEvidenceValidationData supporting or refuting a hypothesisView reference → that it will), and Ease 1-10 (the inverse of cost). Compute composite (I × C × E or (I + C + E) / 3). Rank. Discuss the top 3-5; the rest are noise.
Treating ICE scores as objectiveObjectiveStrategyA strategic goal (OKR)View reference → when they're calibration exercises. Letting the team with the loudest voice anchor the scoring (anonymous scoring, then group discussion, fixes this). Confusing Ease with Effort: ICE inverts effort so that high-ease items get high scores, which makes the composite read in the right direction.
Pair with RICE when the candidates need a Reach multiplier (B2C growth experiments especially). Pair with Build-Measure-Learn to keep the top-ranked candidates running as actual experiments. Use Kano Model before ICE to make sure the candidates being scored span Must-Have, Performance, and Delighter categories, not just one.
Sean Ellis introduced ICE in growth-team practice around 2014 as a simplification of RICE (which Intercom had documented). Ellis founded GrowthHackers and codified ICE as the default ranking heuristic for experiments at the bottom of the growth funnelFunnelGrowthA conversion funnel tracking user progressionView reference →. The framework has since been adopted broadly by growth teams and is supported natively in tools like Notion, Productboard, and Aha!.
Triaging a growth backlog in twenty minutes
A growth team scores three experiments on Impact, Confidence, and Ease from 1 to 10 and multiplies each: a homepage headline test (Impact 6, Confidence 8, Ease 9) lands at 432, a new onboarding email sequence (Impact 8, Confidence 5, Ease 4) at 160, and a pricing-page redesign (Impact 9, Confidence 4, Ease 2) at 72. The headline test wins not because it moves the most but because it is cheap and likely to work, which is exactly what ICE is built to surface when capacity is the constraint.
Knowing when ICE is good enough
An early-stage team without the traffic data to estimate Reach for RICE uses ICE to rank ten candidate experiments in a single stand-up, scoring each idea quickly and shipping the top two the same week. Because the cost of running the smaller tests is lower than the cost of debating them, the deliberately rough scores do their job: the obviously good and obviously weak ideas separate cleanly, and the team upgrades to RICE only once reach starts to matter.