Skip to main content
← All guides
Tool + guide · September 2026 · 7 min read

RICE prioritization

Reach × impact × confidence ÷ effort. Score your list below — and then see the part every other calculator leaves out: how much the estimates would have to be wrong before the order changes.

Short answer

What is the RICE prioritization framework?

A scoring method from Intercom for ordering a backlog: reach times impact times confidence, divided by effort. The score has no units and no meaning on its own — it exists only to sort one list against itself.

The short version

  1. Reach is people per fixed period; impact is a coarse five-point scale (3, 2, 1, 0.5, 0.25); confidence is 100%, 80% or 50%; effort is person-months.
  2. Effort is the divisor, so it has the most leverage — and it is the number people misjudge most, especially for work they want to do.
  3. Confidence multiplies the whole score: a 50% item needs twice the reach to tie an otherwise identical 100% one.
  4. Two scores 3% apart are tied. An order that a single confidence notch reshuffles has not told you what to build.

Free · No signup · Nothing leaves your browser

Score the list, then see how much it would take to reorder it

Reach is people per period — quarter, month, whatever you use, as long as it is the same for every row. Effort is person-months. The score has no units and matters only as an ordering.

ItemReachImpactConfidenceEffort, pm

The order

  1. 1CSV exportscore 500
  2. 2Self-serve onboardingscore 480
  3. 3Dark modescore 125
  4. 4SSO for enterprisescore 64

How solid is that order

  • CSV export over Self-serve onboarding survives a 4% error in reach or effort.

    ⚠️ “CSV export” drops below “Self-serve onboarding” if its impact goes one step down.

    ⚠️ “CSV export” drops below “Self-serve onboarding” if its confidence goes one step down.

    ⚠️ “Self-serve onboarding” overtakes “CSV export” if its impact goes one step up.

    ⚠️ “Self-serve onboarding” overtakes “CSV export” if its confidence goes one step up.

  • Self-serve onboarding over Dark mode survives a 284% error in reach or effort.

  • Dark mode over SSO for enterprise survives a 95% error in reach or effort.

Paste into your roadmap review — fragility included.

What this cannot know. Every number above is an estimate nobody checked. Reach is guessed, impact is a five-point opinion, and effort is underestimated as a rule — which matters more than the rest because effort is a divisor: halve it and the item doubles. The score has no units and no meaning outside this list.

RICE ranks what you already decided to consider. Whether those are the right things to consider is a different question — the Kano model separates features people miss when absent from features they merely enjoy, which is not visible anywhere in a RICE table.

Test the idea behind it

RICE orders a backlog somebody already decided was worth building. A 15-minute session ends with a written GO / WAIT / NO-GO and the reasoning behind it. Free tier, no card.

The four inputs

Reach

people per period

How many people this affects in a fixed window — per quarter, per month, whatever you pick, as long as every row uses the same one. Count people, not events, and count the ones who actually encounter it rather than everyone on the list.

How it goes wrong: The most quietly invented number in the table. “All our users” is almost never the reach of anything.

Impact

3 / 2 / 1 / 0.5 / 0.25

How much it moves the thing you care about, per person reached. Massive, high, medium, low, minimal. The scale is coarse on purpose: it is a judgement, and a finer scale would only make the judgement look measured.

How it goes wrong: Everything becomes “high” when the person who proposed it fills it in.

Confidence

100% / 80% / 50%

How sure you are of the other three. It is the only input that admits you might be wrong, and it multiplies the whole score — so a 50% item needs twice the reach to tie with an otherwise identical 100% one.

How it goes wrong: Set to 100% everywhere, which removes the only brake the framework has.

Effort

person-months

Total work across everyone involved, design and testing included. It is the divisor, which gives it more leverage than any other input: halve the estimate and the item doubles.

How it goes wrong: Underestimated as a rule, and underestimated most for the things people want to build.

The question nobody asks of a ranked list

A sorted list is persuasive in a way the underlying numbers are not. Item one sits above item two, and the eye reads that as a finding — whether the gap is three times or three per cent.

But every input here is an estimate. Reach is guessed from a funnel nobody instrumented. Impact is an opinion on a five-point scale. Effort is underestimated as a matter of professional habit. If the top two items are two per cent apart, the framework has not chosen between them — it has reported the noise in your estimates with impressive precision.

So the calculator above answers the question directly: for each adjacent pair, how large an error in reach or effort the ordering survives, and which single notch on the impact or confidence scale would swap them. Where one notch is enough, the order is a coin landing, and it is better to know that before the quarter is planned around it.

What RICE cannot do

It cannot compare unlike things. A bug fix, a rebrand and a compliance deadline in one table produce a number for each and a decision for none. Rank within a kind, not across kinds.

It cannot see dependencies. The framework treats every item as an independent choice. Work that unblocks other work scores as though it stood alone, and therefore scores badly.

It cannot tell you what belongs on the list. RICE orders the candidates you already chose. Whether they are the right candidates is a different question — the Kano model separates the features people miss when absent from the ones they merely enjoy, and nothing in a RICE table shows that difference.

It cannot make an estimate true. Arithmetic on invented inputs produces confident invented outputs, which is the whole risk of scoring frameworks and the reason the fragility column exists.

Ordering the list assumes the list is worth building

A backlog is a good problem to have. It assumes people want the product and the question is only what to do first. If that assumption is the shaky part, a better ordering of the wrong list does not help.

A 15-minute session works through seven criteria — is the problem real, does a market exist, will they pay — and ends with a written verdict and the reasoning.

Test the idea behind the backlog →

15 min · free tier, no card

Frequently asked questions

What is the RICE prioritization framework?+
A scoring method introduced by Intercom in 2016 for ordering a backlog. Each candidate gets four numbers — reach (people per period), impact (a five-point scale from 3 down to 0.25), confidence (100%, 80% or 50%) and effort in person-months — and the score is reach times impact times confidence, divided by effort. The score has no units; it exists only to produce an ordering.
How do you calculate a RICE score?+
Multiply reach by impact by confidence, then divide by effort. A feature reaching 1,200 people a quarter, with high impact (2), 80% confidence and four person-months of work scores 1200 × 2 × 0.8 ÷ 4 = 480. That 480 means nothing on its own — it is only meaningful next to the scores of the other items you calculated the same way.
Why does effort matter more than the other inputs?+
Because it is the divisor. Doubling reach doubles the score, and halving effort also doubles it — but effort is the input people misjudge most, and they misjudge it hardest for the work they are keenest to do. Anything estimated at under a person-month deserves particular suspicion: the leverage is largest exactly where the estimate is weakest.
How reliable is a RICE ranking?+
Exactly as reliable as the estimates, which is rarely discussed because the output looks like arithmetic. Two items whose scores differ by 3% are, for practical purposes, tied — and a list where one confidence notch reshuffles the top three has not told you what to build. The calculator on this page reports, for each adjacent pair, how much error in reach or effort the ordering survives, and which single change would swap them.
When is RICE the wrong tool?+
When the items are not comparable — a bug fix, a rebrand and a compliance deadline do not belong in one ranking. When the work is a dependency rather than a choice. And when the real question is which kinds of features matter at all, rather than which order to build a settled list in; that question belongs to something like the Kano model.
Is this calculator free, and does my data leave the browser?+
It is free and requires no account. Everything is computed in your browser — nothing you enter is sent to us or stored anywhere.

Related guides