RICE Prioritization Framework: Formula, Scoring Guide and a Worked Example
The formula, the scoring scales, a computed example with a stress test, and an honest look at where RICE misleads.
RICE scores an idea as (Reach × Impact × Confidence) ÷ Effort. Take reach from real data, use a 0.25–3 impact scale, tie confidence to evidence, and measure effort in person-months. Stress-test the top scores, treat small gaps as ties, and use the ranking to start a discussion, not to end one.
RICE is the most quoted prioritization framework in product management and also one of the most misused. Used well, it forces four hard questions into the open and makes different ideas comparable. Used badly, it produces a spreadsheet that looks scientific and simply confirms what the loudest person already wanted. This guide shows how to use it well.
What RICE is
RICE scores each idea on four things and combines them: Reach (how many people it affects in a set time), Impact (how much it moves the needle for each of them), Confidence (how sure you are about the first two) and Effort (how much work it takes). The score is:
RICE score = (Reach × Impact × Confidence) ÷ Effort
Bigger is better. Reach × Impact is the total value, Confidence discounts it for uncertainty, and Effort turns it into value per unit of work.
How to score each part
| Part | How to measure | Typical scale |
|---|---|---|
| Reach | People or events affected per time period, from real data | e.g. 5,000 users per quarter |
| Impact | Effect on each person toward your goal | 3 massive, 2 high, 1 medium, 0.5 low, 0.25 minimal |
| Confidence | How sure you are, based on evidence | 100% strong data, 80% some data, 50% a hunch |
| Effort | Total team time: design, engineering, QA | Person-months |
Two rules keep this honest. First, take reach from analytics, not from imagination: “how many users hit this screen last quarter?” Second, tie confidence to evidence: if your only evidence is a stakeholder’s opinion, the score is 50% or less, no matter how excited people are.
A worked example, computed
Five ideas for a learning app, scored over one quarter (illustrative numbers). The scores below are calculated, not estimated:
| Idea | Reach | Impact | Confidence | Effort | RICE score |
|---|---|---|---|---|---|
| Simplified onboarding | 20,000 | 1 | 100% | 2 | 10,000 |
| Reminder nudges | 12,000 | 0.5 | 80% | 1 | 4,800 |
| Referral rewards | 8,000 | 2 | 80% | 4 | 3,200 |
| Search filters | 5,000 | 2 | 50% | 3 | 1,667 |
| Offline mode | 3,000 | 3 | 50% | 6 | 750 |
Simplified onboarding ranks first (10,000) because it reaches many people, is cheap, and we are confident in it. Offline mode ranks last (750) despite a massive per-user impact: it is expensive and our confidence is only 50%. That is exactly what RICE is for. It stops a big, exciting idea from beating a cheap, broad one just because it sounds more impressive.
Test the ranking before you trust it
Scores are only as good as their inputs, so stress-test them. What if the impact of “Reminder nudges” is 1 instead of 0.5? Its score becomes 9,600, very close to Simplified onboarding. That tells you the top two would be effectively tied, and the decision depends on one guess: how much the nudge really matters. The right move is to test that guess cheaply, not to argue about it. A framework that shows you which assumption matters most is doing its real job.
Where RICE goes wrong
False precision. A score of 3,333 feels exact; it is not. Treat small differences as ties. Gaming. People adjust confidence or reach until their favourite wins. Have someone other than the proposer challenge the inputs. Ignoring strategy. A high score on the wrong goal is still the wrong work. Comparing unlike things. Fixing a security hole and launching a growth feature do not belong on the same scale; some work simply must be done. And RICE does not capture learning: a low-scoring experiment may teach you something that changes every other score.
RICE compared with other methods
| Method | Strength | Weakness | Use when |
|---|---|---|---|
| RICE | Considers reach and confidence | Needs decent data | Comparing many similar ideas |
| ICE | Fast, simple | No reach; more subjective | Quick triage of experiments |
| MoSCoW | Good for fixed releases | No ranking inside categories | Scoping a deadline-bound release |
| Value vs effort grid | Visual, quick | Value is subjective | Workshops with mixed stakeholders |
Estimating each input when you have no data
| Input | If you lack data, try |
|---|---|
| Reach | Use a proxy: how many users reach the screen this touches? Take a similar feature’s usage as a range. Ask analytics for a rough figure rather than guessing. |
| Impact | Compare with a past change you know the result of. “This is about as big as the onboarding fix that lifted activation by 3 points.” |
| Confidence | Count your evidence: user research, data, a competitor’s result. No evidence means 50% at most. |
| Effort | Ask the engineers for a range in person-weeks and the biggest unknown. Never estimate for them. |
Using ranges is more honest than a single number. Score the best case and worst case; if the ranking flips between them, the idea needs a cheap test before it needs a place on the roadmap.
A spreadsheet you can set up in ten minutes
A: Idea B: Problem it solves C: Reach (per quarter) D: Impact (0.25 to 3) E: Confidence (%) F: Effort (person-months) G: Score = C × D × E ÷ F H: Evidence I: Who scored it
Add column H, evidence, and never leave it blank. Sort by G. Then read the top ten with your team and ask, for each: does this ranking match our gut? If not, which input is wrong?
Running a scoring session with your team
- Agree the goal and the time period for reach (for example, one quarter).
- Product proposes scores; design and engineering challenge them.
- Engineering gives effort, not product.
- Anyone can question any input, but must say what they would change it to and why.
- Record disagreements instead of forcing consensus; they point to what to test.
The value of the session is not the final list; it is the twenty minutes of argument about what reach and impact really are. Teams often discover they were assuming very different things.
Using RICE for experiments and technical work
RICE works for experiments if you treat “impact” as the expected size of the learning-based improvement and set confidence low. For technical debt, it is harder: reach may be “engineers affected” and impact “time saved per engineer”, but often it is better to reserve capacity for debt than to compete it against features. Do not force everything into the same formula; use it where it clarifies, and drop it where it distorts.
Common mistakes
Scoring confidence by how much you like the idea. Using reach numbers nobody can source. Letting effort be estimated by the person proposing the feature. Treating the ranking as the decision instead of the start of a conversation. Re-scoring only when a new idea arrives, so old scores go stale. And building a complicated spreadsheet to avoid making a judgment call.
Where RICE fits in the bigger picture: how to prioritize product features, and how the result feeds a roadmap. For interviews, see the execution section of the 50 question bank.
Frequently asked questions
What is the RICE prioritization framework?
A scoring method that ranks ideas by Reach, Impact, Confidence and Effort. The score is Reach times Impact times Confidence, divided by Effort, so higher scores mean more value per unit of work.
How do you calculate a RICE score?
Multiply Reach (people affected in a period), Impact (a 0.25 to 3 scale) and Confidence (a percentage), then divide by Effort in person-months.
What are the limitations of RICE?
It can give false precision, be gamed by adjusting inputs, ignore strategy, and compare unlike work such as security fixes and growth features. Treat scores as input to a discussion, not as the decision.
When should I not use RICE?
When you lack reliable reach data, when work is mandatory (security, regulation), or when comparing fundamentally different kinds of work on one scale.
Want this sharpened for your specific situation?
I take on a small number of career calls and resume reviews each month.
See how we could work together