Ice scoring model

ICE is a prioritization framework that scores ideas, features, or initiatives on three dimensions: Impact, Confidence, and Ease. Each dimension receives a score (typically 1-10), and the scores are multiplied together to produce an overall priority score. Ideas with higher ICE scores are prioritized over those with lower scores.

The three dimensions

Impact measures the potential positive effect if the idea succeeds. This might be measured as expected effect on a key metric, number of users affected, or business value generated. High impact means the idea could meaningfully move outcomes that matter.

Confidence represents how certain you are about your Impact and Ease estimates. Low confidence acknowledges uncertainty-you're guessing based on limited data. High confidence means you have evidence supporting your estimates: user research, comparable past results, or experimental data.

Ease captures how simple the idea is to implement. This encompasses effort required, resources needed, technical complexity, and risks. Higher ease means quicker, cheaper, less risky implementation.

How ice scoring works

For each idea being prioritized:

The multiplication means that ideas need to score reasonably across all dimensions. A highly impactful but low-confidence idea will have a moderate score. An easy but low-impact idea won't rise to the top.

Example ice scoring

The login bug, despite moderate impact, ranks highest because it has high confidence and ease. The search rebuild, despite high impact and confidence, ranks third because low ease drags down the score.

When ice works well

ICE is particularly useful for:

Quick prioritization. When you need to rank many ideas without extensive analysis, ICE provides structure without heavy overhead.

Growth experiments. Teams running frequent experiments benefit from ICE's simplicity and the explicit acknowledgment of confidence levels.

Cross-functional alignment. The three dimensions create a common framework for discussing trade-offs between product, engineering, and business perspectives.

Early-stage products. When data is limited and many directions are possible, ICE's inclusion of confidence prevents over-weighting uncertain opportunities.

Ice limitations

Subjectivity. All three dimensions involve judgment. Different people might score the same idea quite differently. The scores feel precise but reflect opinions.

Multiplication quirks. Very low scores in any dimension can dominate. An idea with Impact 10, Confidence 2, Ease 10 scores 200, lower than Impact 5, Confidence 5, Ease 5 scoring 125. Whether this reflects appropriate penalties is debatable.

Missing dimensions. ICE doesn't explicitly capture factors like strategic alignment, dependencies, or opportunity cost. Important considerations may be overlooked.

Scale inconsistency. What does Impact "7" mean? Without calibration, scores aren't comparable across teams or time periods.

Gaming potential. Teams motivated to advance certain ideas may inflate scores. Without accountability, ICE becomes a rationalization tool.

Ice vs. rice

RICE (Reach, Impact, Confidence, Effort) is a similar framework with key differences:

RICE may be more appropriate when Reach varies significantly across ideas. ICE may be simpler when audience size is relatively constant.

Making ice more effective

Define scales. Create explicit definitions for what each score level means. What does Impact 8 look like? What evidence supports Confidence 6?

Score collaboratively. When multiple people score independently and then discuss differences, the conversation often matters more than the final number.

Update confidence. After implementing ideas, compare actual results to predictions. Over time, this improves estimation accuracy.

Combine with other inputs. ICE provides one perspective. Strategic considerations, user feedback, and technical dependencies should also inform prioritization.

Avoid false precision. The difference between 336 and 340 is noise. Use ICE for rough ranking, not precise ordering.

Tools like Klero help improve ICE scoring by grounding estimates in real customer feedback. When you can see how many users requested a feature and how intensely, Impact and Confidence estimates become more evidence-based and less speculative.