RICE Prioritization: Formula, Example and Pitfalls

RICE scoring ranks features by multiplying Reach, Impact and Confidence, then dividing by Effort. Reach is how many people a feature affects in a set period, Impact is how much it moves each of them on a fixed scale, Confidence is a percentage for how sure you are of the first two, and Effort is the work in person-months. The highest score is the most value per unit of work. It is a good way to put a long list of features in a defensible order, and a poor way to settle a question the inputs cannot answer. Below are the formula, a worked example, a spreadsheet and the pitfalls.

If you have not yet decided whether RICE is the right framework at all, feature prioritization frameworks compares it with MoSCoW, Kano and WSJF.

The RICE formula

RICE score = (Reach x Impact x Confidence) / Effort

The method comes from Intercom. Sean McBride’s post on RICE prioritization↗ (opens in a new tab) defines each factor and gives fixed scales so that different people score the same way:

• Reach: how many people or events the feature affects in a defined period, such as customers per quarter. Use real numbers from your product data where you can.
• Impact: how much the feature moves each person toward the goal. Massive is 3, high is 2, medium is 1, low is 0.5 and minimal is 0.25.
• Confidence: how well supported your reach and impact estimates are. High is 100%, medium is 80% and low is 50%; anything lower, the post says, is a “total moonshot.”
• Effort: the total work from everyone involved, in person-months, estimated in whole numbers or at least half a month.

The score reads as “total impact per time worked.” Its units do not matter, only the order. A score of 600 is not twice as good as 300 in any real sense; it just ranks higher.

A worked RICE example

An invented example: a small software company with about 4,000 active customers and five candidate features for next quarter. Reach is customers per quarter, effort is person-months.

Feature                    Reach  Impact  Conf.  Effort  Score
-------------------------  -----  ------  -----  ------  -----
Slack notifications         1,500    0.5    80%     1      600
New onboarding checklist    2,000    2      50%     4      500
Dark mode                   3,000    0.25   80%     1.5    400
Bulk CSV import               600    1      80%     2      240
Single sign-on (SSO)          120    3     100%     3      120

Read the result as a first draft, then question it:

• Slack notifications win because they are cheap and reach a lot of people, even with low impact each. That is RICE working as intended.
• The onboarding checklist has the largest total impact but only 50% confidence. The useful move is not to build it or drop it, but to raise the confidence cheaply, for example with five customer interviews or a prototype test, and rescore.
• Dark mode scores well on reach alone. Check whether “reach” here means people who would use it or people who merely see a settings page.
• SSO comes last, yet three prospective enterprise customers will not sign without it. That is a table-stakes feature, and RICE is the wrong tool for it; override the score and say why.

A RICE spreadsheet in two minutes

Intercom offers its own spreadsheet from the post, but building one takes two minutes and makes the formula yours. Put one feature per row, the four inputs in columns B to E, and two formulas beside them. Both Google Sheets and Excel have RANK.EQ↗ (opens in a new tab), which gives tied scores the same rank.

     A                 B       C        D           E        F                      G
1    Feature           Reach   Impact   Confidence  Effort   Score                  Rank
2    Slack notifs      1500    0.5      80%         1        =B2*C2*D2/E2           =RANK.EQ(F2,$F$2:$F$6)
3    Onboarding        2000    2        50%         4        =B3*C3*D3/E3           =RANK.EQ(F3,$F$2:$F$6)
...
Add columns H "Reach source" and I "Override reason".
Data validation on C: 3, 2, 1, 0.5, 0.25. On D: 100%, 80%, 50%.

Two details make the sheet trustworthy. Restrict Impact and Confidence to the fixed values with a dropdown, so nobody scores impact 2.7. And add a “Reach source” column that says where each reach number came from, such as “analytics, July to September” or “guess.” Microsoft’s help on RANK.EQ in Excel↗ (opens in a new tab) describes the same function for teams on Excel.

RICE pitfalls

1. False precision. Four guesses multiplied together produce a number with three significant figures. Treat scores within about 20 percent of each other as a tie and break it with judgment.
2. Mixed reach units. One feature measured in customers per quarter and another in page views per month cannot be compared. Fix the unit and period for the whole sheet.
3. Confidence that never drops. If every row says 100%, the column does nothing. A rule of thumb: no data behind the estimate means 50% at most.
4. Effort that only counts engineering. The Intercom post counts time from product, design and engineering. Documentation, support training and migration belong in the estimate too.
5. Ignoring dependencies. A low-scoring piece of plumbing may be what three high scorers need. Score the group, or override with a reason.
6. Table stakes. Must-have features, what the Kano model calls must-be quality↗ (opens in a new tab), earn no praise and so score low on impact, yet customers leave without them. Handle them outside the score.
7. Scoring to win. Once people learn the formula, reach estimates for their own ideas creep up. Have one person own the inputs, and ask for the source of every reach number.
8. Scoring once. Reach and confidence change as you learn. Rescore the top of the list each month, not the whole sheet.

McBride is direct about the limit: “RICE scores shouldn’t be used as a hard and fast rule.” His point is that when you do override the score, for a dependency, a table-stakes feature or strategy, the score makes the trade-off visible. That is the real value of RICE scoring: not the number, but the written reason whenever you do not follow it.

RICE on a fenbs board

fenbs has no score field and no custom fields, so the arithmetic stays in your sheet and the board holds the result. A pattern that works:

• Each candidate is a task of kind feature or enhancement. The note holds the problem and the evidence for reach, such as who asked and how often.
• The RICE inputs and score go in a comment, for example “RICE: R 1,500/qtr, I 0.5, C 80%, E 1, score 600.” When an estimate changes, add a new comment, so the old one stays on record.
• Map the ranked order to priority, 1 to 10 with 1 the most urgent: the top two or three become P1 or P2, the next handful P3 to P5.
• The optional size, XS to XL, is a quick visual cue for effort, not a replacement for person-months.
• Record the method as a rule on the Decisions and rules page: the reach period, the effort unit and when overrides are allowed. Connected AI assistants read the rules first.

An AI assistant connected over MCP can do the tedious part. Ask it: “Read the RICE comments on the open features, list them in score order with any missing inputs, and propose a priority for each. Change nothing.” You decide, and every priority you change is recorded in History with your name.