How to score a backlog with RICE without gaming the numbers
A seven-step guide to scoring a backlog with RICE, keeping the numbers honest, and publishing a ranked list with the reasons behind each rank.

Every backlog has the same problem. Each stakeholder argues for their own item, and the loudest voice tends to win. RICE gives you one way to compare items. It only works if the numbers are honest, and numbers are easy to bend. This guide shows how to score with RICE and keep the scores straight.
In brief: who it is for
Product managers with a backlog where every stakeholder argues for their own item.
What you will have at the end
A ranked list and the reasons behind each rank, ready to publish.
RICE came from Intercom. The formula is short. The hard part is that each of the four inputs needs a rule, agreed before anyone scores anything. Without the rules, the formula just turns opinions into decimals.
The four inputs and their rules
| Input | What it measures | The rule |
|---|---|---|
| Reach | How many people the item affects in a fixed period, say a quarter | Counted from data, not guessed |
| Impact | How much it moves one goal | Fixed scale: 3 massive, 2 high, 1 medium, 0.5 low, 0.25 minimal |
| Confidence | How sure you are of the other numbers | 100% high, 80% medium, 50% low. Below 50% it is a research task, not a build |
| Effort | Work needed, in person-months | Across every discipline: design, engineering, QA. Estimated with engineering, never alone |

The seven steps
Agree the period and the one goal
Do this before you score anything. Pick a fixed period, such as a quarter, and one goal. Every impact score will be measured against that goal and no other. If stakeholders want a second goal, run a second scoring round.
Score reach from data
Count the people the item would affect in the period. Use your product data. If you cannot count it, write down what you would need to count it. A guess dressed as a number is how arguments restart later.
Score impact against the goal
Use the fixed scale and nothing in between: 3, 2, 1, 0.5 or 0.25. Ask one question for each item: how much does this move the goal for the people it reaches? Not how much does it matter in general.
Set confidence and write the evidence
Choose 100%, 80% or 50%. Beside the number, write what you are basing it on: a test, interviews, usage data, or a hunch. If the honest answer is below 50%, stop. That item is a research task, not a build.
Estimate effort with engineering
Count person-months across design, engineering and QA. Do the estimate with engineering in the room. An estimate made alone tends to leave out the work you cannot see.
Compute, sort, then check the top five
Multiply, divide, sort. Then read the top five out loud. Ask whether the order feels right, and if it does not, find which input is off. Do not nudge the final score. You can run this in a spreadsheet or in PikoPM's Prioritisation tool.
Publish the scores with the reasons
Share the ranked list with the evidence beside each number. A stakeholder whose item ranked low can see why, and can bring better evidence next round. That is a better conversation than a lobbying session.
How RICE gets gamed
Three things go wrong again and again.
Impact is inflated. Everyone's item becomes a 3. Confidence stays at 100% by default. Nobody changes the field, so nothing is ever uncertain. Effort is low-balled. A small number makes the score jump, and the real cost shows up later.
The fixes are plain. None needs a new tool.
Score together, in one room. A score defended in front of design, engineering and the other stakeholders is harder to inflate than one typed alone.
Write the evidence beside every confidence number. It makes a default of 100% visible, and uncomfortable.
Re-score items after they ship. Compare the estimates with what happened. You will see whose estimates were right, and the room will know you will check.
A worked example
The numbers below are illustrative. They are not from a real product. The goal is: more trial users activate. The period is one quarter.
| Item | Reach | Impact | Confidence | Effort | Score |
|---|---|---|---|---|---|
| A. Onboarding checklist | 4,000 | 2 | 80% | 2 | 3,200 |
| C. Dark mode | 6,000 | 0.25 | 80% | 1 | 1,200 |
| B. Single sign-on | 300 | 3 | 100% | 3 | 300 |

The order is A, C, B. The onboarding checklist wins because it reaches many people, moves the goal, and costs two person-months. Dark mode reaches the most people, but its impact on activation is minimal, so it ranks second. Single sign-on has the highest impact per person, yet it reaches 300 people and costs three person-months.
Look at B's confidence. It sits at 100%. Under the rules above, that needs evidence written beside it. If the evidence is thin, the score drops and B falls further.
Now the part that matters. B may be a must-have for one large deal. RICE does not know that. It is an input to judgement, not the decision. If you overrule the order and build B first, say so out loud and write down why. Then the list stays honest, and the exception is on the record where the team can see it.
Before you publish
- The period and the single goal are written at the top of the list.
- Every reach number comes from data.
- Impact uses only the fixed scale.
- Every confidence number has its evidence beside it, and nothing below 50% is in the build list.
- Effort was estimated with engineering and covers every discipline.
- The top five were read out loud and checked.
- Any overruled item is marked, with the reason.
A scored backlog will not end every argument. It moves the argument to the evidence, where it belongs. Re-score after shipping, and your team's estimates get better each quarter.


