THE ANSWER
ICE scoring rates each growth idea from 1 to 10 for impact, confidence and ease, then multiplies the three. Run the highest scores first and re-score as you learn.
Deb Pradhan · Founder, 100xLab
Published 5 MIN READ
ON THIS PAGE · 8
ICE scoring is a quick way to rank growth ideas. You rate each one from 1 to 10 for impact, confidence and ease, multiply the three numbers, and run the highest scores first. It takes minutes, and it replaces "whoever argues best wins" with a list the team can read.
This post gives the formula, a scoring guide you can copy, a worked example with made-up ideas, and the mistakes that make ICE scores useless.
What is ICE scoring?#
ICE is a prioritisation model for experiments and growth ideas. The three letters stand for:
- Impact: how much the idea could move the number you care about, if it works.
- Confidence: how sure you are that it will have that impact.
- Ease: how little time, money and effort it needs.
The model is usually credited to Sean Ellis, who coined the term "growth hacking", and Morgan Brown, who wrote about it in the book Hacking Growth. ProductPlan's glossary (opens in a new tab) describes it as a way to rank ideas by those three factors. It was designed as a "good enough" estimate, not a precise calculation, which is why it is quick to use.
What is the ICE scoring formula?#
The most common version is:
ICE score = Impact × Confidence × Ease
Each factor gets a score from 1 to 10. The product runs from 1 to 1,000. Higher is better.
Sources differ on the detail. Some guides multiply the three, others average them. We multiply, because it punishes an idea that is weak on any single factor. An idea that is easy and sure but has tiny impact should not rank high, and multiplication makes sure it does not. Whichever method you pick, use it for every idea.
How do you score impact, confidence and ease?#
The scores only work if everyone means the same thing by a 3 or an 8. Write the scale down before you start. Here is a guide you can copy:
| Score | Impact | Confidence | Ease |
|---|---|---|---|
| 1 to 3 | Small move in the target number | A hunch, no evidence | Weeks of work or a new hire |
| 4 to 6 | Noticeable move | Some evidence, such as a competitor doing it | A few days of one person's time |
| 7 to 8 | Large move | Our own data points the same way | A day or two |
| 9 to 10 | Changes the quarter | Tested before, same result | A few hours |
Three rules keep the scores honest.
Tie impact to one number. Pick a single target, such as qualified leads per month, and score every idea against it. "Brand awareness" is not a number you can compare with a conversion rate.
Base confidence on evidence. A past test, your analytics, customer calls and a clear reason all count. "I just feel good about this" is a 2. This is why tracking comes first: without it, nearly every confidence score is a guess. Our post on analytics before ads covers the event plan that gives you something to point at.
Score ease as the whole cost. Include design, build, review and tracking, not just the first draft.
What does a worked example look like?#
These ideas and scores are made up for illustration. They show how the method behaves, not what any company should expect.
Imagine a seed-stage software company. Its target number is booked demos per month. The team scores five ideas.
| Idea | Impact | Confidence | Ease | ICE |
|---|---|---|---|---|
| Rewrite the pricing page to answer the top three sales objections | 7 | 7 | 8 | 392 |
| Add a demo-booking button to every blog post | 5 | 6 | 9 | 270 |
| Launch a comparison page against the main competitor | 7 | 5 | 6 | 210 |
| Start a weekly webinar | 6 | 3 | 3 | 54 |
| Rebuild the whole site on a new design | 8 | 4 | 1 | 32 |
The pricing page rewrite wins, even though the rebuild has the biggest impact on paper. The rebuild is slow and the team has little evidence it would work, so its score collapses. That is the point of the model: it stops a big, exciting, uncertain project from beating a small project that will certainly help.
Note what the ranking does not say. It does not say the webinar is a bad idea forever. It says that, today, with this evidence, it is a poor bet.
How do you run an ICE session?#
Here is a process that takes about an hour.
- Set one target. Write the number and the date. For example, "booked demos per month, by the end of the quarter".
- List the ideas. Collect 10 to 20 from the team, from customer calls and from your analytics. Write each as an action, not a theme.
- Agree the scale. Paste the scoring guide above into the document. Change the wording to fit your business.
- Score alone first. Each person scores every idea privately. This stops the loudest voice setting the number.
- Compare and discuss the gaps. Where scores differ by three points or more, talk. The gap usually hides a fact one person knows and another does not.
- Rank and pick two or three. Take the top scores, not the top ten. A short list gets finished.
- Write the test. For each pick, record the hypothesis, the number you will watch and the date you will read it.
- Re-score after the results. Results change confidence. Update the list and pick the next two or three.
The last step matters most. ICE is a loop, not a one-off ranking. For a longer process around it, see how we set up growth work from diagnosis to a 30-day results check.
What are the limits of ICE scoring?#
ICE is quick, and the price of quick is subjectivity. Be aware of four problems.
Scores drift upward. Teams like their own ideas, so confidence creeps up. Counter it by requiring a piece of evidence for any confidence score above 6.
It ignores reach. A change on a page nobody visits can score well on ease and impact. If you have traffic data, check how many people an idea touches. RICE, the version that adds reach, exists for this reason.
It rewards the easy. Because ease is multiplied in, quick wins crowd out slow, important work. Keep one slot per cycle for a high-impact idea with a low ease score, and plan it separately.
It does not set the target. ICE ranks ideas against a number. If the number is wrong, the ranking is wrong too. Pick the target first. If you are working out which metric matters, our customer acquisition cost guide is a good starting point for the cost side.
ICE, RICE or PIE: which should you use?#
| Model | Factors | Best for |
|---|---|---|
| ICE | Impact, confidence, ease | Fast ranking with little data |
| RICE | Reach, impact, confidence, effort | Product and growth work where you have reach numbers |
| PIE | Potential, importance, ease | Ranking pages for optimisation |
For an early-stage team, ICE is enough. You rarely have clean reach data, and a ten-minute scoring pass beats a long debate. Move to RICE when you have the traffic and funnel numbers to make reach a fact rather than a guess.
Where does 100xLab fit?#
ICE tells you what to try first. Someone still has to build it, which is the work of marketing engineering. 100xLab builds marketing as systems at a fixed price, with analytics set up before anything else and a results check at 30 days, so your next ICE round starts from numbers rather than hunches. See what we build or how Sprint, Launch and Partner are priced. We do not run your ad budget or your outbound.
SERVICE
GROWTH & ANALYTICS
Know where buyers drop off, and what fixes it.
FAQ
05 QUESTIONS
01What does ICE stand for?
Impact, confidence and ease. Impact is how far the idea could move your target number, confidence is how sure you are, and ease is how little effort it needs.
02Do you multiply or average the ICE scores?
Most guides multiply the three scores, which punishes an idea that is weak on any one of them. Some teams average instead. Pick one method and use it for every idea, so the ranking stays comparable.
03What is the difference between ICE and RICE?
RICE adds reach, the number of people an idea touches, and divides by effort. ICE is lighter and faster. Use ICE when you need a quick ranking, and RICE when you have reach data you trust.
04How many ideas should we score at once?
Score a list of 10 to 20 ideas, run the top two or three, then score again. A long backlog goes stale, and results from the first tests change your confidence in the rest.
05Is ICE scoring reliable?
It is a structured guess, not a measurement. Its value is that it makes the team state its assumptions and compare ideas on the same terms. Use data to set the scores wherever you have it.