RICE, ICE and the limits of a score
Scoring frameworks are good at making you ask the right questions and bad at answering them for you.
Nº 16 · HeedPost team · Sep 22, 2026 · 2 min read
Sooner or later every product team discovers scoring frameworks. RICE, from Intercom, scores an idea by Reach, Impact, Confidence and Effort. ICE, popularised by Sean Ellis for growth experiments, uses Impact, Confidence and Ease. There are dozens of variations. Put numbers in, get a ranked list out.
They're genuinely useful. They're also easy to misuse in a way that looks rigorous and isn't.
What they're good at
A scoring framework forces you to think about things you'd otherwise skip. How many people will this actually reach? How sure are we that it'll help? What does it really cost? Teams that never asked those questions make better decisions the first time they do, regardless of the arithmetic.
Frameworks also make disagreements visible. If one person scores impact as 3 and another as 1, you've found the conversation you needed to have.
Where they go wrong
False precision. Impact and confidence are guesses. Multiplying guesses together produces a number with two decimal places and no more truth than the guesses. A score of 41.7 beating 38.2 means nothing.
Gaming. Once people know the formula, the numbers drift towards whatever gets their idea built. Confidence goes up when someone really wants something.
Missing strategy. A framework ranks ideas against each other. It can't tell you that the top three all pull the product in a direction you don't want to go.
Feed them real inputs where you can
The weakest part of most scores is reach and impact, because they're usually estimated in a meeting. Your feedback board can do better than an estimate:
- Reach: how many distinct people asked, after merging duplicates.
- Impact: who those people are and what they're worth to you, not how loudly they asked.
- Confidence: how many of them you've actually talked to.
Effort still needs your engineers. There's no shortcut for that, and there shouldn't be.
Use the score to start the conversation
Our suggestion: score your top twenty requests roughly, in a few minutes each, then throw away the exact numbers and look at the order. Does it surprise you? If yes, why? That discussion is the real output. The ranked list is just a way to have it.
And keep the score off your public board. Customers don't need to know a request scored 41.7. They need to know whether it's planned and why.