ICE Scoring (Impact, Confidence, Ease)

ICE scoring is a lightweight way to decide what to work on first. The letters stand for Impact, Confidence, and Ease. For each idea on a list, you give a quick score — typically one to ten — on three questions. Impact: if this works, how much will it move the metric you care about? Confidence: how sure are you that it actually will, given the evidence you have? Ease: how cheap and quick is it to try, in time and resources? Multiply the three numbers and you get an ICE score, and sorting by that score ranks the backlog. The method was popularized by Sean Ellis, who coined "growth hacking" and used ICE to help teams choose among the flood of experiments they could run. Its whole appeal is speed: a team can score dozens of ideas in under an hour and walk out with an ordered list.

The reason ICE caught on is that it fights two real problems at once. The first is paralysis — too many possible experiments and no agreed way to choose, so debate eats the week. The second is the seductive pull of big, exciting ideas that are also slow and risky. By forcing Ease and Confidence into the same score as Impact, ICE keeps a flashy long-shot from automatically jumping the queue ahead of a smaller, near-certain, quick win. A growth team staring at forty ideas can score them in a meeting, see that three high-Ease, high-Confidence tests rise to the top, and ship those this week instead of arguing. The number is rough on purpose. ICE is a conversation accelerator, not a precision instrument, and treating it as the latter is the main way teams misuse it.

ICE sits at the simple, fast end of a spectrum of prioritization frameworks, and knowing where it fits keeps you from over- or under-trusting it. Heavier models like RICE add Reach — how many users an idea touches — and divide by Effort, producing a more complete but slower estimate. ICE deliberately leaves Reach out, which is its biggest blind spot: a high-Impact change on a tiny, low-traffic page can score the same as one on your busiest page, even though the second helps far more people. If reach varies wildly across your ideas, a model that accounts for it will rank better than ICE. The trade is honesty about scope versus speed of scoring, and you pick based on how different your ideas' reach really is.

ICE's other well-known weakness is subjectivity. The three scores are estimates, and two reasonable people will often score the same idea differently, especially on Impact and Confidence where evidence is thin. That is not fatal, but it means ICE works best as a shared, transparent exercise rather than a private calculation: score together, say why, and let the discussion — not the decimal point — do the work. Compared with elaborate weighted models, ICE buys you speed and a common language at the cost of precision and reach-awareness. Use it when you have many small-to-medium ideas and need to choose quickly. Reach for RICE or a fuller model when ideas differ enormously in scope and a wrong call would be expensive. The frameworks are tools for different jobs, not rivals.

Get the most from ICE by treating it as a team ritual, not a spreadsheet truth. Agree in advance what Impact is measured against — which single metric — so everyone scores the same thing. Score live, together, and make people justify outlier numbers out loud, because the disagreement is where the learning is. Keep the scale consistent across sessions so scores stay comparable over time. Then act on the ranking quickly and, crucially, record what actually happened: real results feed back into your Confidence scores for similar future ideas, so the method gets smarter as you run it. The output you want is not a perfect ordering but a fast, shared, defensible decision about what to ship next — and a habit of revisiting the list as evidence comes in.

The pitfalls are easy to name. The biggest is false precision — reading a score of 168 as meaningfully better than 162 when both are guesses; ICE sorts ideas into rough tiers, not exact ranks. The second is gaming, where someone inflates Confidence to push a pet idea up the list, which honest, public scoring discourages. The third is forgetting that ICE ignores Reach, so it can over-rank a high-impact tweak that barely anyone will see. The fourth is scoring once and never updating Confidence with what you learn, so the model never improves. Avoid them by scoring transparently as a team, treating scores as tiers, watching for the reach blind spot, and feeding real outcomes back in. Done that way, ICE turns an endless backlog debate into a quick, repeatable decision.

ICE scoring — Impact times Confidence times Ease — is a fast prioritization method popularized by Sean Ellis for ranking growth experiments quickly.

Etymology: source.

Search interest for this term over the last five years:

View interest-over-time on Google Trends →

Areas of marketing where ice scoring (impact, confidence, ease) is a core concern: