RICE Prioritization Framework: Score & Rank Your Backlog (2026)

RICE is a prioritization framework that ranks ideas by computing a single score from four inputs: Reach, Impact, Confidence, and Effort. The formula is (Reach × Impact × Confidence) / Effort, and the output is a ranked list defensible enough to share in a planning meeting without an hour of debate.

Looking for the calculator with worked examples? See RICE score calculator: the formula with 3 worked examples.

RICE was published by Sean McBride at Intercom in 2016 as the framework Intercom's product team actually used. The format caught on because it converts the soft "this feels important" conversation into specific, comparable numbers — without pretending to be more precise than it is. (If you're not sure RICE is even the right tool for your decision, start with how to pick a framework in the first place.)

Score your backlog on your phone — Framework for iPhone & iPad computes RICE scores with AI-assisted inputs. Free to start.

The four inputs

• Reach — how many people the change affects in a defined time period. Customers per quarter, signups per month, daily active users touched per week. Pick the unit once and use it for every idea in the comparison.
• Impact — how much it changes things per person when it lands. Intercom uses a 5-point scale: 3 = massive, 2 = high, 1 = medium, 0.5 = low, 0.25 = minimal. The bins are deliberately coarse to prevent fake precision.
• Confidence — how sure you are about the Reach and Impact estimates. 100% = strong evidence, 80% = some evidence, 50% = gut feel. Anything below 50% means the score is mostly noise.
• Effort — total person-months to ship. Sum across roles. Use the same unit for every idea.

The score is (Reach × Impact × Confidence) / Effort. Higher = better.

How to actually use it

1. List 10–30 ideas. Don't bother below 10; the ranking is meaningless with too few. Don't go above 30 in one session; the team gets fatigued and starts rubber-stamping.
2. Score Reach with concrete numbers, not relative scales. "Affects 8,000 users per quarter" beats "high reach." Look up the actual numbers from analytics rather than guessing.
3. Score Impact discreetly per idea, not relatively. Don't peek at other ideas' impact scores until you've scored this one independently. Relative scoring drifts toward the middle.
4. Score Confidence honestly. If you've never tested anything like this, you're at 50%, not 90%. The whole point of the C term is to penalize beautiful-but-speculative ideas.
5. Estimate Effort with two people. Solo estimates are systematically optimistic. Pair-estimating cuts the systematic bias in half.
6. Compute the score, then read the ranked list. Don't tweak the inputs to make a favored idea win — that's the failure mode RICE was designed to prevent.

What RICE is good at

• Cross-functional planning meetings where engineering, design, and product are arguing about which feature to build next. RICE gives everyone the same vocabulary.
• Quarterly roadmap selection when you have more ideas than capacity. The Effort denominator forces honest cost accounting.
• Documenting the why of a prioritization decision for a stakeholder who wasn't in the room. The spreadsheet is the artifact.

What RICE is bad at

• Highly novel work where Confidence is below 50% for everything. The output is too noisy to be useful — use a premortem instead to surface what you don't know.
• Strategic positioning ("should we enter this market"). Use a SWOT or Porter's Five Forces at that level; RICE assumes the market is already chosen.
• Ideas with externalities — a project that unblocks 10 other projects has a Reach RICE can't capture in one number. Add a side column for dependencies and read both.

The "I cheated the scores" failure mode

The most common abuse of RICE: someone has an idea they want to win, and they inflate Reach or Impact until the score comes out right. Two practices defend against this:

• Calibrate against a known winner. Pick one feature you already shipped and know was successful. Score it with RICE. Every new idea has to be compared to that baseline — Reach and Impact in the same units, scaled to the same scale.
• Separate the scoring from the deciding. The person who advocates for an idea should not be the same person who scores it. Get a second set of eyes on Reach, Impact, and Confidence before computing.

If a team can't score the same ideas the same way twice, the team doesn't yet have the shared context to use RICE — and that's a real signal worth listening to.

When to graduate from RICE

Once a team has used RICE for 4–6 quarters and has a real history of "we scored this X and the actual outcome was Y," the team will know which terms it systematically over- or under-estimates. At that point, replace RICE with a custom scoring rubric that reflects what the team has actually learned. RICE is the training-wheels version; the customized rubric is the bicycle.

The RICE Confidence Discount

The single most common failure mode in real RICE practice is Confidence inflation: teams default the C input to 0.80 on every item, which neutralizes the math because everything gets the same multiplier. The Confidence Discount is a named correction: every quarter, score the actual delivery outcomes of last quarter's RICE-ranked items, and compute a team-specific Confidence multiplier from the gap between predicted and actual Impact × Reach.

The protocol:

1. Pull last quarter's RICE-scored items that actually shipped.
2. Re-score Impact and Reach with hindsight. What did each item actually deliver?
3. Compute the ratio of actual-to-predicted (Impact × Reach) for each item.
4. Take the median ratio across all items. This is your team's Confidence Discount.
5. Multiply next quarter's Confidence inputs by this discount. A team scoring everything at 0.80 might find their median ratio is 0.75 — meaning their true Confidence baseline is closer to 0.60, and items they think are "80% confident" are actually behaving like "60% confident reporting at 80% out of habit."

A worked re-calibration on a 5-item shipped batch:

Item	Predicted Impact × Reach	Actual	Ratio
Onboarding redesign	1.0 × 8,000 = 8,000	0.5 × 5,000 = 2,500	0.31
Search filter v2	0.5 × 12,000 = 6,000	0.5 × 9,000 = 4,500	0.75
Notification opt-in	0.25 × 20,000 = 5,000	0.25 × 18,000 = 4,500	0.90
AI-assistant launch	2.0 × 4,000 = 8,000	1.0 × 3,500 = 3,500	0.44
Mobile checkout fix	1.0 × 15,000 = 15,000	1.0 × 14,000 = 14,000	0.93

Median ratio: 0.75. The team's stated Confidence numbers should be multiplied by 0.75 next quarter. Items the team scores at 0.80 actually behave like 0.60. The ranking shifts accordingly — items that depend heavily on Impact estimates (which historically miss harder) drop relative to items dominated by Reach (which historically lands closer). After 2-3 quarters of re-calibration, most teams converge on a stable team-specific Confidence baseline somewhere in the 0.55-0.70 range.

Why this works: it makes Confidence an empirical input rather than a self-report. The biggest wins from RICE come not from the formula itself but from teams learning which kinds of items they systematically over-confident on. Confidence ratios near 0.30-0.40 (the AI-assistant launch row above) flag categories of work where the team is consistently wrong — a structural learning, not just a single-item correction.

Compare the alternatives at RICE vs ICE (when Reach is too hard to estimate at all) and RICE vs MoSCoW (when scoring itself is too heavy for the team's stage). Teams running SAFe often reach for WSJF (Weighted Shortest Job First) instead, which divides cost of delay by job size rather than weighing Reach and Impact separately — see the full head-to-head at RICE vs WSJF.

Related frameworks

• Pareto Analysis (80/20 Principle) — for spotting the 20% of items that drive 80% of value before scoring everything in the backlog
• RICE vs MoSCoW — when to bucket scope vs when to score
• RICE vs ICE — the Reach term, and when dropping it is the right call
• ICE Score — the lighter cousin (Impact × Confidence × Ease), useful when Reach is hard to estimate
• MoSCoW — categorical alternative when scoring is too heavy
• Weighted decision matrix — for choosing between options rather than ranking features
• Kano Model — pair with RICE when classifying feature types matters more than score
• RICE for SaaS teams — industry playbook: tuning Reach and Confidence for recurring-revenue roadmaps
• RICE for agency teams — industry playbook: scoring client work when Effort is billable hours

Want to try it? Open the RICE entry → for the catalog template, or sign in to use it in the canvas with your real backlog.