ICE Scoring:How to Prioritize Growth Experiments
ICE scoring ranks growth experiments by Impact × Confidence ÷ Ease, each scored 1–10. How to use it to pick the one experiment worth running this week.
ICE scoring prioritizes growth experiments on three dimensions — Impact, Confidence, Ease — each scored one to ten, combined as Impact × Confidence ÷ Ease. You run the highest-scoring experiments first. The framework exists because the experiment you feel most strongly about is usually the one least worth running.
Here's how each dimension is scored, why the formula divides instead of multiplies on the third one, and why the boring experiments keep winning.
The problem ICE scoring solves
Six months after launch, the situation is always the same. You have somewhere between twenty and a hundred customers. There are roughly fifteen things you could work on next week, and you have time for exactly two. The other thirteen wait, possibly forever.
Most founders pick using three signals: what's most fun, what feels most urgent, and what their loudest customer most recently asked for. None of these are bad signals — but none of them are good signals either, and the difference between the founder who grows steadily over the next twelve months and the founder who plateaus is almost entirely whether they let those three signals run the roadmap.
ICE replaces the three vibe signals with a score. It takes about ten minutes to score a backlog of fifteen ideas, and the ranking it produces will disagree with your gut more often than you expect. That disagreement is the value.
The three ICE dimensions, scored one to ten
| Dimension | The question | How to score it |
|---|---|---|
| Impact | If this works, how much does it move the metric? | Size of the movement × importance of the metric. A signup headline change might move signup conversion 20%; a pricing page change might move revenue per signup 10%; an email sequence change might move activation 40% |
| Confidence | How sure are you it will work? | 1 = wild guess, 10 = "I've heard this from twenty customers." Evidence from customer conversations and similar changes elsewhere raises it. Most experiments live in the 4–7 range |
| Ease | How fast and cheap is it to run? | Lower is easier. A headline change is a 1 (fast, cheap). A new onboarding flow is a 6 (a week to build). A pricing model overhaul is a 10 (a month, with reversibility risks) |
Two of the dimensions are about the upside; one is about the cost. That's why the formula treats them differently.
The formula: Impact × Confidence ÷ Ease
The ICE score for an experiment is Impact multiplied by Confidence, divided by Ease. The highest-ICE experiments are the ones with high impact, high confidence, and low cost — and because Ease sits in the denominator, cheap experiments get systematically promoted.
That's deliberate. Cheap experiments let you run more experiments, and the compounding value of an experiment programme comes from volume of validated learning, not from any single result. A founder who runs ten small, well-supported tests learns more in a quarter than a founder who spends the same quarter on one heroic pricing overhaul that may not even be reversible.
The bookThe Solopreneur RevolutionLuis Gonçalves on building a solopreneur business with AI while employed.Get the book →Why your favourite experiment is usually the wrong one
ICE sounds bureaucratic, and most founders' first instinct is to skip the scoring and just pick what feels right. The reason to actually do the scoring is that it forces you to compare experiments on the same dimensions — and that comparison exposes an uncomfortable truth: the experiment you feel most strongly about is often the lowest-ICE one on the list.
High impact in your imagination. Low confidence in reality. High cost to run. That's the anatomy of the pet project — the redesign, the bold new feature, the pricing revolution. It scores badly on every dimension except emotional attachment, and emotional attachment isn't a dimension.
The experiments that actually move the business are usually the boring high-ICE ones: small changes, well supported by customer evidence, fast to run. Nobody brags about them. They compound anyway. The discipline of scoring is what makes you actually choose them over the exciting loser.
ICE inside your weekly experiment rhythm
ICE is a prioritization layer, not a process on its own. It plugs into the experiment loop — hypothesis, test, measure, learn, one experiment per week — covered in full in how to run growth experiments. The combined workflow looks like this:
- Keep a backlog. Every idea that comes from your Monday dashboard review, a customer conversation, or a churn email goes on the list. Nothing gets built on arrival.
- Score new ideas as they land. Thirty seconds each: Impact, Confidence, Ease, one to ten.
- Run the top of the list. One experiment at a time, so you can attribute the result cleanly.
- Re-score after every learning. A finished experiment changes your confidence in related ideas — the write-up from one signup-friction test raises the Confidence score of every other friction-removal idea on the backlog.
A solo founder can run maybe twenty to thirty real experiments per year. That's your entire budget for learning what grows the business — which is exactly why each slot has to go to the highest-ICE candidate, not the loudest one. Scoring is how a one-person company makes prioritization decisions a growth team would respect. It's also the muscle the Solopreneur Accelerator builds in the growth weeks: backlog, scores, one clean test at a time.
12 weeksYour corporate career was never the planApply to the Solopreneur Accelerator.Apply now →Frequently asked questions
What does ICE stand for in growth?
Impact, Confidence, Ease. Impact is how much the change would move the metric if it works; Confidence is how sure you are it will work; Ease is how fast and cheap it is to run. Each is scored one to ten, and the combined score ranks your experiment backlog.
How is the ICE score calculated?
Impact × Confidence ÷ Ease. Ease is scored with lower meaning easier — a headline change is a 1, a new onboarding flow a 6, a pricing overhaul a 10 — so dividing by it rewards cheap, fast experiments that let you run more tests per year.
What is a realistic confidence score?
Most experiments live in the four-to-seven range. A one is a wild guess; a ten means something like "I've heard this from twenty customers." Very-high-confidence experiments are rare — if everything on your backlog scores nine, you're scoring your enthusiasm, not your evidence.
Is ICE scoring overkill for a solo founder?
The opposite. A solo founder gets maybe twenty to thirty real experiments a year, so a wasted slot costs proportionally more than it does at a startup with a growth team. Ten minutes of scoring protects fifty weeks of testing — it's the cheapest insurance in the whole growth stack.
How trapped are you, really?
Take the Corporate Suffocation Index, the free solopreneur readiness assessment.