RICE Prioritization: A Worked Example, Sensitivity Check, and Blank Template

Use the score to expose assumptions before choosing work

RICE combines reach, impact, confidence, and effort into one comparison: reach × impact × confidence ÷ effort. Intercom's original explanation defines reach over a shared period, impact per affected person, confidence in the estimates, and total team effort. It also allows a team to depart from score order when dependencies or other commitments justify doing so.

The score is a decision aid, not a forecast of revenue, completed tasks, or customer happiness. A precise-looking result can still come from weak assumptions. The useful part of the exercise is making those assumptions visible enough that someone can challenge them.

This guide works through three fictional improvements to a library room-booking service. It includes the first calculation, a corrected reach estimate, the resulting decision, and a reusable blank brief.

Choose one goal, one period, and one unit of reach

The fictional team wants to improve independent booking completion during the next quarter. It compares three discretionary improvements: a clearer booking summary, an optional reminder, and additional search filters. Essential accessibility repairs, active incidents, and binding commitments are handled as obligations before this comparison; they are not made optional by receiving a low score.

Reach is the number of distinct eligible people who would encounter the proposed change during the quarter. That definition rules out mixing annual visitors for one idea with monthly sessions for another. It also stops a frequent user from being counted twenty times for one idea and once for the next.

Impact is a shared team rating of the expected contribution to the booking-completion goal for an affected person. In this example, 2 means high, 1 medium, and 0.5 low. These ratings are not measured percentage-point changes. Confidence is entered as a fraction, so 80% becomes 0.8. Effort uses person-weeks across the whole team; this is an explicit local unit choice rather than the person-months in Intercom's example.

Write the evidence beside each input

For the booking summary, the example team estimates 600 people will reach the relevant step. It has repeated observations of confusion about the reservation details, so it assigns high impact, confidence 0.8, and two person-weeks of effort. Those are still estimates: observations identify a problem but do not prove the proposed design will solve it.

For reminders, the team estimates 900 eligible people, medium impact, confidence 0.5, and three person-weeks. The weaker confidence reflects uncertainty about whether the reminder will affect the chosen goal. A reminder might be valuable for attendance while doing little for booking completion. If attendance is the actual goal, it needs its own coherent comparison.

For filters, the first pass uses 1,500 people, low impact, confidence 0.8, and one person-week. This row will initially rank highest. The evidence note matters because it reveals that 1,500 came from all search users, not necessarily people who would encounter or use the new filter. The arithmetic cannot detect that scope mismatch on its own.

Calculate the first ranking and read its units honestly

A scores 600 × 2 × 0.8 ÷ 2 = 480. B scores 900 × 1 × 0.5 ÷ 3 = 150. C scores 1,500 × 0.5 × 0.8 ÷ 1 = 600. The initial order is C, A, B. It is a ranking under these stated assumptions, not a commitment to deliver in that order.

Do not translate 480 into 480 additional completed bookings. Impact is a judgment rating, confidence is an estimate adjustment, and effort is a planning denominator. The resulting score combines those chosen scales. It does not inherit the meaning of a measured product outcome.

Also do not add the reach of A, B, and C to announce 3,000 unique beneficiaries. People may encounter more than one improvement. RICE scores for separate ideas do not automatically model overlap, dependencies, or diminishing returns when a team selects a portfolio.

Test the assumption that can reverse the decision

The team checks the filter exposure definition before scheduling the work. In the fictional corrected count, only 700 distinct people meet the relevant conditions and would encounter the filter during the quarter. Keeping the other inputs fixed gives 700 × 0.5 × 0.8 ÷ 1 = 280. A now ranks above C.

The break-even reach for C is easy to calculate. To equal A's score of 480, C needs R × 0.5 × 0.8 ÷ 1 = 480, so R = 1,200. This tells the team why the reach question matters. If a credible range for C is 600–900 people, its score is 240–360 and A stays ahead under the other unchanged assumptions.

Now test A's effort. If a dependency increases it from two to four person-weeks, A's score falls to 240. C at the corrected reach would then score 280 and move ahead. The decision is therefore sensitive to both exposure and delivery effort. The team should clarify those two inputs before spending time debating a tiny difference in an arbitrary impact rating.

Turn the ranking into a decision with a review point

In the worked scenario, the team confirms the corrected filter reach of 700 and keeps A's two-person-week estimate after a brief dependency check. It selects the booking-summary improvement first. The decision note records the inputs and explains why the earlier filter-first ranking changed.

The team does not treat the confidence number as an 80% probability of success. It records what would change its mind: a materially larger delivery estimate, a scope change, or evidence that the proposed summary does not resolve the observed confusion. A prototype check can test that last assumption before full implementation.

After release, compare the actual relevant journey with the expected change using an appropriate measurement design. A before-and-after count alone may be affected by seasonality, traffic mix, or other releases. Keep the prioritization estimate separate from the evidence about what the delivered work actually achieved.

Copy this brief before adding another idea

A useful template captures the decision context and the source of each estimate, not just four numbers. Fill it before the ranking meeting so disagreements can focus on the same question. If an input is unknown, record the unknown and the smallest useful check instead of inventing precision.

Keep scopes comparable. A two-day copy improvement and a six-month platform replacement may not belong in the same near-term decision unless the team has defined what it can actually choose. If an idea is a dependency, name the dependent work and its necessity; do not hide it by inflating impact until the score wins.

Finish with a decision owner and a reason to revisit. The framework has done its job when the team can explain why it chose the work, which assumptions mattered, and what evidence would change the choice.

References

• Intercom: Original RICE Prioritization Explanation