An ICE score is a quick way to rank ideas by rating each one from 1 to 10 on three factors, Impact, Confidence and Ease, and combining the three ratings into one number. The original GrowthHackers method averages them: ICE score = (Impact + Confidence + Ease) ÷ 3. Many teams multiply them instead (Impact × Confidence × Ease). Either way, the ideas at the top of the sorted list are the ones to test first.
ICE scoring was created by Sean Ellis, who coined the term “growth hacking” and founded GrowthHackers, to compare growth experiment ideas without long debates. This guide covers how to score each factor, when to average and when to multiply, worked examples, a copyable ICE scoring template, how ICE compares with RICE, and the mistakes that make ICE scores unreliable.
What Is an ICE Score?
ICE is a prioritization framework for experiments and small projects. Each idea gets three ratings on a 1 to 10 scale, where higher is always better. GrowthHackers’ own documentation for its NorthStar product defines them this way (GrowthHackers):
- Impact: the impact the experiment will have on your team’s goals.
- Confidence: how confident you are that the experiment will work.
- Ease: the effort and resources the experiment needs, scored so that easier ideas get higher numbers.
In a Breakout Growth podcast episode, Ellis explains the trade-off ICE is built to expose: the ideal idea is high impact, takes almost no effort and is very likely to work, but those ideas are rare. Most of the time a high-impact idea is hard to run, or an easy idea is unlikely to move much. ICE makes those trade-offs visible on one scale.
Ellis and Morgan Brown also describe ICE in their 2017 book Hacking Growth, and product coach Itamar Gilad later popularized a version of ICE for product ideas, with a structured way to rate confidence (more on that below).
The ICE Score Formula: Average or Multiply?
There are two common ways to combine the three ratings:
| Method | Formula | Score range | Who uses it |
|---|---|---|---|
| Average | (Impact + Confidence + Ease) ÷ 3 | 1 to 10 | GrowthHackers’ NorthStar product averages the three ratings (source) |
| Multiply | Impact × Confidence × Ease | 1 to 1,000 | Itamar Gilad’s ICE guide uses ICE score = Impact × Confidence × Ease (source) |
The difference matters when one rating is very low. Averaging lets a strong impact score cover for weak confidence. Multiplying punishes any single low rating, so a long-shot idea can’t float to the top on impact alone. Neither is wrong. Pick one method, write it at the top of your sheet and use it for every idea, because scores from the two methods can’t be compared.
ICE score (average) = (Impact + Confidence + Ease) / 3
ICE score (multiply) = Impact × Confidence × Ease
How to Score Impact, Confidence and Ease
ICE uses relative 1 to 10 ratings, so each team decides what the numbers mean. Gilad’s advice is that any definition works as long as the team applies it consistently. The anchors below are a starting point you can adapt, not an official scale.
Impact: how much will it move the goal?
Score impact against one goal per scoring round, such as signups, activation, conversion rate or retention. Ellis suggests looking at a model of the business: if a change could double a key step in your main funnel, it may deserve a high score, while a 10% lift on a step few people reach probably won’t move the needle.
| Impact | Suggested meaning |
|---|---|
| 9 to 10 | Could change the goal metric dramatically, for example doubling a core funnel step |
| 6 to 8 | A clear, noticeable lift on the goal metric |
| 3 to 5 | A modest lift, or a big lift on a small segment |
| 1 to 2 | Barely measurable on the goal metric |
Confidence: how strong is your evidence?
Confidence is where ICE scores go wrong most often, because people score their excitement instead of their evidence. Ellis points out that qualitative research, such as customer interviews and surveys, and results from similar past tests are good reasons to raise confidence. Gilad goes further with his Confidence Meter, which assigns confidence according to the type of evidence you have. In his scale, your own conviction, industry trends and colleagues’ opinions add almost nothing, while customer research, test results and launch data add much more.
| Confidence | Suggested meaning |
|---|---|
| 9 to 10 | Results from a test or launch of this same idea |
| 6 to 8 | Data from similar past experiments, or strong customer research |
| 3 to 5 | Some supporting data, such as survey results or a few customer requests |
| 1 to 2 | A hunch or an opinion, with no data yet |
Ease: how little time and money does it take?
Ease is the inverse of effort: the less work an idea needs, the higher the score. Agree on a conversion table, such as person-days or person-weeks per score, so two people scoring the same idea land on the same number. Ellis also notes that ease often rewards creativity: a smaller “minimum viable test” of the same hypothesis can turn a hard idea into an easy one.
| Ease | Suggested meaning |
|---|---|
| 9 to 10 | Less than a day of work, no engineering |
| 6 to 8 | A few days, one or two people |
| 3 to 5 | One to three weeks, several people |
| 1 to 2 | A month or more, or a major build |
ICE Scoring Examples
The examples below are illustrative. The ideas and ratings are invented to show the math for a growth team whose goal is trial signups. They are not benchmarks.
| Idea | Impact | Confidence | Ease | ICE (average) | ICE (multiply) |
|---|---|---|---|---|---|
| Add customer review snippets to the pricing page | 6 | 7 | 9 | 7.3 | 378 |
| Exit-intent signup offer on blog posts | 4 | 6 | 8 | 6.0 | 192 |
| Rebuild the onboarding flow | 9 | 5 | 3 | 5.7 | 135 |
| Launch a referral program | 8 | 4 | 4 | 5.3 | 128 |
Here the two methods agree: the review snippets win because they’re easy and backed by decent evidence, and the referral program comes last because it’s hard and unproven. The math for the first row: (6 + 7 + 9) ÷ 3 = 7.3, or 6 × 7 × 9 = 378.
When averaging and multiplying disagree
Now compare two more illustrative ideas:
| Idea | Impact | Confidence | Ease | ICE (average) | ICE (multiply) |
|---|---|---|---|---|---|
| Idea A: bold new homepage headline nobody has tested | 10 | 2 | 9 | 7.0 | 180 |
| Idea B: add a comparison table to the top product page | 6 | 6 | 6 | 6.0 | 216 |
Averaging puts Idea A first, because its impact and ease outweigh its very low confidence. Multiplying puts Idea B first, because a confidence of 2 drags A’s product down. If your team tends to over-rate exciting ideas, multiplying is the safer choice.
ICE Score Template
Copy this table into a spreadsheet, replace the sample rows with your own ideas and sort by your chosen ICE column. The sample rows are placeholders.
| Idea | Goal metric | Impact (1-10) | Confidence (1-10) | Ease (1-10) | ICE (average) | ICE (multiply) | Evidence / notes |
|---|---|---|---|---|---|---|---|
| Example: review snippets on pricing page | Trial signups | 6 | 7 | 9 | 7.3 | 378 | Similar test on another page won |
| Example: rebuild onboarding flow | Activation | 9 | 5 | 3 | 5.7 | 135 | Survey data only |
| Your idea | |||||||
| Your idea |
Google Sheets version
To build the template in Google Sheets or Excel, paste this into cell A1. It’s tab-separated, so each value lands in its own column. Columns F and G calculate both versions of the score for row 2:
Idea Goal metric Impact (1-10) Confidence (1-10) Ease (1-10) ICE (average) ICE (multiply) Evidence / notes
Example idea Trial signups 6 7 9 =IFERROR(ROUND(AVERAGE(C2:E2),1),"") =IFERROR(C2*D2*E2,"")
Drag both formulas down, then delete the column you don’t use so nobody mixes methods. Two setup tips:
- Add dropdowns for the ratings. Use Data > Data validation with a dropdown of 1 to 10 on columns C to E, so nobody types 0 or 15.
- Make the notes column required. A short note on where each confidence score came from (“past test,” “survey,” “hunch”) keeps scores honest and easy to challenge.
How to Run an ICE Scoring Session
- Set one goal. Agree on the metric you’re scoring impact against for this round.
- Collect ideas before scoring. List every candidate first so early favorites don’t anchor the rest.
- Score one factor at a time across all ideas. Rate impact for every idea, then confidence, then ease. Comparing down a column keeps the scale consistent.
- Have the idea’s owner score it, then review as a group. Discuss only the scores that look off, not every number.
- Sort and pick the top few. Test the highest scores, and add a short hypothesis for each test.
- Rescore after each test. Compare actual results with the impact you predicted. Over time your team’s scoring gets sharper.
ICE vs RICE
The RICE framework, created at Intercom, adds a separate Reach factor and measures effort in person-months instead of a 1 to 10 ease rating. Its formula is (Reach × Impact × Confidence) ÷ Effort.
| ICE | RICE | |
|---|---|---|
| Factors | Impact, Confidence, Ease | Reach, Impact, Confidence, Effort |
| Scales | 1 to 10 for each factor | Reach as a count, impact 0.25 to 3, confidence as a %, effort in person-months |
| Formula | Average or product of the three ratings | (Reach × Impact × Confidence) ÷ Effort |
| Speed | Very fast; minutes per idea | Slower; needs reach data and effort estimates |
| Best for | Growth experiment backlogs where reach is similar across ideas | Roadmaps where ideas affect very different numbers of people |
Use ICE when you need to rank a long list of quick tests every week. Switch to RICE when reach varies a lot between ideas, for example a change to one landing page versus a change every user sees, because ICE folds reach into the impact guess.
Common ICE Scoring Mistakes
- Scoring enthusiasm as confidence. A popular idea with no data is still low confidence. Ask what evidence backs the number.
- Mixing average and multiply. Scores from the two methods aren’t comparable. Choose one for the whole backlog.
- No shared definitions. If one person’s 7 is another person’s 4, the ranking is noise. Write the anchors at the top of the sheet.
- Scoring impact against different goals. Keep one goal metric per scoring round.
- Ignoring reach. ICE has no reach factor, so an idea that touches few users can score as high as one that touches everyone. Factor reach into impact or use RICE.
- Treating the score as the decision. ICE is a relative ranking to start a conversation. Dependencies and strategy can still justify going out of order.
- Never rescoring. Scores go stale. Update them when tests return results.
ICE pairs well with customer metrics. If you track loyalty with an NPS score, detractor comments are a good source of experiment ideas, and the score can be the goal metric you rate impact against.
Frequently Asked Questions
What does ICE stand for?
ICE stands for Impact, Confidence and Ease. Each idea is rated from 1 to 10 on all three, and the ratings are combined into one ICE score used to rank ideas.
How do you calculate an ICE score?
Rate Impact, Confidence and Ease from 1 to 10, then average them ((I + C + E) ÷ 3, the GrowthHackers method) or multiply them (I × C × E, a widely used variant). For ratings of 6, 7 and 9, the average is 7.3 and the product is 378.
Who created the ICE scoring model?
Sean Ellis, the founder of GrowthHackers who coined the term “growth hacking,” created ICE scoring to prioritize growth experiments. He and Morgan Brown describe it in the 2017 book Hacking Growth.
Should you average or multiply ICE scores?
Either works if you use one method consistently. Averaging (1 to 10) is the original GrowthHackers approach. Multiplying (1 to 1,000) penalizes any single low rating more, which helps if your team tends to over-rate exciting but unproven ideas.
What is a good ICE score?
There’s no universal good ICE score. Scores depend on how your team defines each rating, so they only mean something relative to other ideas scored the same way. Work from the top of your sorted list.
What is the difference between ICE and RICE?
RICE adds a separate Reach factor and measures effort in person-months, using the formula (Reach × Impact × Confidence) ÷ Effort. ICE is faster and uses 1 to 10 ratings; RICE is better when ideas reach very different numbers of people.
Can you use ICE scoring outside growth teams?
Yes. Product, marketing, SEO and content teams use ICE to rank backlogs. Keep the same rules: one goal per scoring round, shared definitions for each rating and a notes column for evidence.
Can AI engines read your site?
Check any page for free: crawler access for ChatGPT, Perplexity, Claude and Google AI Overviews, plus the fixes that matter most.


