ICE Scoring for Growth Ideas

Growth teams usually have more ideas than time. A single brainstorm can produce forty suggestions, and a small SaaS team might realistically run three or four experiments a month. ICE scoring for growth ideas is a simple way to decide which ones to try first. Each idea is rated on three factors, Impact, Confidence, and Ease, and the combined score gives you a rough ranking.

ICE is not precise, and it was never meant to be. Its value is that it makes prioritization conversations more structured and less dominated by whoever argues loudest. This article explains each factor, how to calculate and use the score, how to score well as a team, where the method breaks down, and how to keep a backlog that gets smarter over time.

The three factors

Impact

Impact asks how much the idea will move the target metric if it works. The phrase "if it works" matters. Impact is not a prediction of success; it is a measure of how big the win would be assuming it happens. That separation keeps impact from being mixed up with confidence, which is a common source of muddled scores.

Most teams use a scale from 1 to 10. A score of 1 means a minimal effect that you might not even be able to detect. A 5 means a noticeable improvement. A 10 means a major change to the metric. Impact depends heavily on where in the funnel an idea sits and how many people it touches. A change to the signup flow affects every new user, while a change to an advanced settings page affects only a few.

Confidence

Confidence asks how sure you are that the idea will work. This is where evidence comes in. A score of 1 is a pure guess, an idea that sounded good in a meeting with nothing behind it. A 5 suggests some supporting evidence, such as user feedback, a pattern in analytics, or a well-established practice. A 10 means strong evidence, such as a previous successful test of a very similar idea on your own product.

Confidence is the factor teams most often inflate. People tend to feel confident about their own ideas, so it helps to ask what specifically supports the score. If the answer is "it just makes sense," the score should probably be low.

Ease

Ease asks how easy the idea is to implement. A 1 means weeks of work across several teams. A 5 means a few days of work. A 10 means something that can be done in an hour, such as changing a headline or adjusting an email's send time. Ease should include everything needed to run the experiment, not just to build the change: design, engineering, copy, tracking, and any coordination with other teams.

Calculating the score

There are two common ways to combine the factors. The first is to average them, adding Impact, Confidence, and Ease and dividing by three. The second is to multiply them. Multiplication spreads scores further apart, which can make differences clearer, and it punishes ideas that score very low on any single factor. An idea with high impact and high ease but a confidence of 1 will look reasonable when averaged and much weaker when multiplied, which is often the more honest result.

Either method works. What matters is choosing one and using it consistently, so that scores from different weeks and different people can be compared.

A worked example (illustrative)

Imagine a small SaaS team focused on improving activation, with four ideas on the table. Removing the phone number field from the signup form scores 5 on impact, 7 on confidence, and 9 on ease, for an average of 7.0. The impact is moderate because it only affects one step, the confidence is fairly high because the field is known to cause hesitation, and the change takes minutes.

Redesigning the pricing page scores 8 on impact, 5 on confidence, and 3 on ease, averaging 5.3. It could matter a lot, but the team is unsure what the redesign should change and it will take weeks. Adding an onboarding email for users who stall after signup scores 6, 6, and 7, averaging 6.3. Launching a referral program scores 7, 3, and 2, averaging 4.0, since the team has no evidence its users refer others and the build is substantial.

In this example, removing the phone field and adding the stalled-user email rank highest, so the team runs those first. The pricing page redesign stays on the list, and the team might look for a cheaper way to raise confidence, such as interviewing a few recent buyers, before committing to the full project.

How to score well

Define the metric first

Impact only makes sense relative to a goal. "Will this help growth?" is too vague to score. "Will this increase the share of new signups who complete setup in their first week?" is something you can reason about. Agree on the target metric before scoring, and if the team has several goals, score ideas separately for each or pick one focus per cycle. An idea that would transform demo requests might do nothing for activation, and a single impact score cannot capture both.

Base confidence on evidence

Ground confidence in something you can point to. Analytics data can show where users drop off. User research can show why. Past experiments on your own product are the strongest evidence of all. Results from similar companies can help too, though they transfer less reliably than people expect, because audiences and products differ. When someone gives an idea high confidence, ask them to name the evidence in a sentence and record it next to the score.

Estimate ease with the people doing the work

The person proposing an idea is often not the person who will build it, and they tend to underestimate effort. Ask engineering, design, and marketing to estimate ease for the parts they would own. This takes a few extra minutes and prevents the common situation where a "quick win" turns out to need a database change and two weeks of work.

Score as a group

Scoring together reduces individual bias. A practical way to do it is to have each person score independently and then compare. Where scores are close, take the average and move on. Where they differ widely, discuss the difference. Large disagreements are often the most useful part of the exercise, because they reveal that people have different information or different assumptions about what the idea involves.

Common pitfalls

Scores that feel arbitrary

ICE is subjective, and a score of 6 versus 7 rarely reflects a real difference. Teams that treat the scores as exact sometimes argue over decimals while ignoring the bigger picture. Use ICE to structure discussion and to sort a long list into rough tiers, not to make decisions automatically. If the top few ideas are close, judgment should decide among them.

Overweighting ease

Easy ideas tend to dominate an ICE list, because ease is the factor people can estimate most confidently and it often scores high. Over time, a team that only runs the top of the list can end up running many small tweaks and never attempting the bigger changes that might matter more. One fix is to reserve some capacity each cycle for a high-impact idea, even if its total score is lower, and to break big ideas into smaller tests that raise confidence before full investment.

Inconsistent scoring

Different people use scales differently. One person's 7 is another's 4. A short scoring guide with descriptions for low, middle, and high scores on each factor, ideally with an example from your own backlog, keeps everyone closer to the same scale. Revisit the guide when new people join.

Never revisiting scores

Scores go stale. A new feature can make an idea easier, a competitor's move can change its likely impact, and a related experiment can raise or lower confidence. Review the scores of ideas that have sat in the backlog for a while rather than assuming they are still accurate.

Alternatives to ICE

RICE

RICE adds Reach to Impact, Confidence, and Effort. Reach estimates how many users an idea will affect in a given period, which helps compare ideas that touch very different numbers of people. It is a good fit for product teams prioritizing features, and it pushes teams to think explicitly about scale instead of folding it into impact.

PIE

PIE scores ideas on Potential, Importance, and Ease, and is common in conversion optimization, where teams are choosing which pages to test. Importance usually reflects how valuable the traffic to a page is.

A simple impact-effort matrix

Some teams skip numbers entirely and plot ideas on a two-by-two grid of impact and effort. High-impact, low-effort ideas go first, and low-impact, high-effort ideas are dropped. This is fast and visual, and it works well for small teams with short lists.

Choose the framework that fits how your team works. Consistency matters more than the specific method, since any of them is better than prioritizing by whoever spoke last.

Building a growth backlog

Keep ideas in a single backlog that records, for each one, the idea itself, the hypothesis behind it, the metric it targets, its ICE scores with a note on the evidence for confidence, its status, and its owner. A spreadsheet is enough for most early-stage teams.

The hypothesis is worth writing in full. "If we remove the phone field, more visitors will complete signup, because the field creates hesitation" makes it clear what you expect and why, which makes the result easier to interpret later. Review the backlog weekly, pick the top ideas, and run experiments. Archive ideas that have been rejected or superseded so the list stays manageable.

Learning from experiments

After each experiment, update the confidence scores for similar ideas based on what you learned. If a test of social proof on the signup page made no difference, lower confidence for other social proof ideas. If a lifecycle email worked well, raise confidence for related email ideas. Over time, your scoring becomes more accurate because it reflects real learning about your product and users rather than general assumptions.

The takeaway

ICE scoring is a lightweight way to bring order to a long list of growth ideas. Define the metric first, base confidence on evidence, estimate ease with the people who will do the work, and score as a group. Use the scores to focus discussion and sort ideas into tiers, revisit them as you learn, and keep room for bigger bets so the list does not fill up with small tweaks.