ICE, and its one weakness
ICE scores each idea on Impact (how much it moves the metric), Confidence (how sure you are it will work) and Effort (what it costs to run), then computes Impact × Confidence ÷ Effort. Higher is better.
The weakness is that the scores are guesses, and guesses drift toward whatever the person scoring already wanted to do. Two habits fix most of it:
- Score confidence on evidence, not enthusiasm. A 9 should mean you have seen this work — in your own data, or at a comparable company. If the only support is that it sounds right, it is a 4.
- Score effort in calendar time, not developer days. An experiment needing design, engineering and legal review is high effort even if the code is small.
How to read the ranking
- Above 15 — run these first. High leverage relative to cost.
- 5 to 15 — worth scheduling once the top of the list is clear.
- Below 5 — drop, or find a cheaper way to test the same hypothesis.
If nothing scores above 15, that is a useful result in itself: your backlog is made of expensive bets. Before committing a quarter, look for a cheaper version of the highest-impact idea that tests the same assumption.
Questions
ICE or RICE?
RICE adds Reach, which matters when features touch very different-sized audiences. For growth experiments at an early stage, reach is usually folded into impact and ICE is quicker to run.
Everything scores about the same. What now?
Your scoring range is too narrow. Force a spread: pick the single highest-impact idea and call it 10, the lowest 1, and place the rest between them.
Does this replace a roadmap?
No. It orders a backlog of experiments. Strategic bets that must happen regardless of score belong on a roadmap, not in a prioritisation formula.
How often should I re-score?
Every time you finish an experiment. Results change your confidence on related ideas, and the ranking should move with them.