ICE score

An ICE score is a quick prioritization number for ideas in a growth or product backlog, built from three ratings: impact, confidence and ease. You rate each idea on all three, combine the ratings into one score, and sort the list so the ideas with the best mix of payoff, certainty and low effort rise to the top.

How it is calculated

Each idea gets a rating from 0 to 10 on impact, confidence and ease, according to Savio. Savio multiplies the three: ICE Score = Impact * Confidence * Ease. Growth Method averages them: ICE Score = (Impact + Confidence + Ease) / 3.

Both versions sort ideas the same basic way, but they produce numbers on different scales. Pick one and use it for the whole backlog.

Worked example

Growth Method's table rates a homepage headline rewrite based on customer interviews at 8 impact, 7 confidence and 9 ease, for a score of 8.0. A referral program with a double-sided incentive gets 9, 5 and 4, for 6.0.

The headline goes first this week. The referral idea has high impact but low ease and confidence. Treat those low ratings as the problem to solve: scope a smaller first version using the referral program playbook, then score it again.

Common mistakes

Reading the score as a verdict. Savio argues that no ICE score is good or bad in itself, and that a high score does not guarantee an idea will work. Growth Method treats 7 or above as high priority on its averaged scale. If you use a threshold, apply it only within your own team's scale.

Optimistic estimates. Savio points to the planning fallacy: teams overestimate value and underestimate effort. Once results come in, compare them with your earlier ratings and adjust how you score.

Scoring pet ideas higher. Bias in the ratings carries straight into the final score. Growth Method suggests defining what each rating means before your first scoring session to prevent anchoring. For example, decide in advance what an impact of 8 means for your signup numbers.

When it matters

ICE helps when you have more experiment ideas than weeks left in the quarter. Score the list in a growth experiment backlog, sort it, then apply judgment. Savio recommends exactly this order: scores first, judgment second.