When Your Prioritisation Framework Lies
RICE, ICE and value versus effort each fail in a specific, predictable way.
You built the spreadsheet. Reach, Impact, Confidence, Effort, one row per idea, sorted descending. The top of the list is three small settings changes and a copy tweak. The thing you believe is the most important work of the year sits eleventh. Everyone looks at the sheet, nobody wants to be the person who overrides the numbers, and the quarter fills with small wins.
The frameworks are fine. They are arithmetic, and arithmetic does not lie. What lies is the inputs, and each framework has a characteristic way of getting them wrong.
Why the framework keeps picking the small thing
How each one fails
- RICE
- Reach times Impact times Confidence, divided by Effort. Effort is the only input you can estimate with any accuracy, so it dominates the ranking. The result is a systematic bias toward small work. It also treats items as independent, which kills anything whose value only appears when three pieces exist together.
- ICE
- Impact, Confidence, Ease, usually scored one to ten from the gut. Fast and honest about being a gut call, which is its virtue in growth experiments. Its failure is that the same person scores all three, so a pet idea quietly gets a ten, a nine and an eight.
- Value versus effort
- The two by two. It is the best tool for a room full of people who disagree, because you argue about position rather than about a number. Its failure is that value is undefined, so each person reads their own definition into the axis and you leave the room agreeing on a picture and disagreeing on the plan.
Worked example
Hypothetical: Fernbank, a savings app
Two candidates. First, a tooltip on the transfer screen: reach 40,000 users a quarter, impact 0.25, confidence 0.8, effort 0.5 person months. RICE: A scoring model: reach, impact, confidence, divided by effort. score is 40000 times 0.25 times 0.8 divided by 0.5, which comes to 16,000. Second, a joint account feature: reach 4,000 users a quarter, impact 2, confidence 0.4, effort 6 person months. That scores 4000 times 2 times 0.4 divided by 6, which is about 533. The tooltip wins by thirty to one. But the tooltip's ceiling is the tooltip. The joint account, if it works, changes who the product is for and brings a second person into every account it touches. RICE has no way to represent a ceiling, so a low variance item with a small upside beats a high variance item with a large one every single time.
That example points at the real issue. These frameworks compute expected value, and expected value is the right criterion when you are making the same bet many times. It is the wrong criterion when a single outcome changes the shape of the business. A portfolio of tooltips never adds up to a joint account.
Where scoring works
- Many similar sized candidates
- Reach numbers you can actually query
- Independent items with no shared unlock
- Reversible, cheap work
- You want to defend a cut line to stakeholders
Where scoring misleads
- One large bet against many small ones
- Reach for a segment that does not exist yet
- Items that only pay off as a set
- Architectural or irreversible choices
- The strategy question is which market you serve
Two fixes that cost you nothing. First, score Confidence separately and never multiply it into a single number you then sort by. Confidence is not value, it is the reason to go and find out. An item with high impact and low confidence is a research task, not a low priority build. Collapsing them hides that.
Second, run the scoring inside a strategy, not instead of one. Decide the guiding policy first, throw out everything that does not serve it, and only then rank what is left. Prioritisation frameworks are good at ordering a set. They are terrible at choosing which set you are ordering, and teams reach for them precisely when they have avoided that choice.
One tell that you are using a framework as an excuse: the output surprises nobody and satisfies nobody, and the meeting ends with someone saying the numbers say we should do the tooltip. Numbers do not say things. People choose inputs.
Quick check
Why does RICE systematically favour small work?
The takeaway
Scoring frameworks rank a set well and choose a set badly, and RICE in particular is biased toward small, safe work.
Try this tomorrow
Re-rank your current backlog with Effort removed entirely. Note which three items moved most, and ask whether their effort is a real reason to skip them.
Answer the check above, then bank the day.
Where this comes from
- RICE: Simple prioritization for product managers, Sean McBride, Intercom
- Escaping the Build Trap, Melissa Perri
- Good Strategy Bad Strategy, Richard Rumelt
Strategy is one of six tracks. These lessons summarise and build on the work above, they do not reproduce it. Buy the books, they are better.