River
Y CombinatorBacked by Y Combinator
FREE TEMPLATE

RICE Prioritization Scoring Template

Every RICE template hands you a grid and a total. This one tells you which single estimate would have to change to reorder the whole list.

Free download  ·  No account needed

Every prioritization template solves the part that was never hard. It gives you columns for reach, impact, confidence and effort, multiplies them, and sorts. Then the order goes out, and about an hour later somebody senior asks why their thing is not first. At that point the grid has nothing left to offer. The only answer available is that the spreadsheet says so, which is an answer that loses the argument and takes the spreadsheet down with it.

So this pack computes the thing that answers the question. Sensitivity Check takes each item, holds three inputs still, and finds the smallest single change to the fourth that would move it up a rank. Then it labels the item Fragile, Contested or Firm. Fragile means the position rests on one judgement call, and you should say so. Firm means somebody would have to halve an effort estimate or double a measured reach figure, which is a much better thing to tell a requester than a rank.

The reason so many items come out fragile is arithmetic rather than sloppiness. Impact is an ordinal scale that gets multiplied, and its notches are doublings: minimal to low, low to medium and medium to high each double the score. Confidence, across its entire range from low to high, also doubles it. The least defensible input in the formula has the same leverage as the whole of the most defensible one, and no template built on this scale mentions it.

The order, and how much of it is actually arithmetic

Every figure in Sensitivity Check recomputes from Scored Backlog. Nothing on it was typed in by a person.

Scored Backlog

Illustrative rows for a fictional field service company, Halyard Dispatch. Reach times impact times confidence, divided by effort.

RankRequestReachImpConfEffScoreImpact basis
1SMS appointment reminders12,0001100%34,000Medium. Cuts no-shows, which cost dispatch time
2Two-factor authentication8,8000.5100%22,200Low. Unblocks procurement, not usage
3Photo attachments on jobs5,200180%22,080Medium. Replaces a text-message workaround
4Bulk technician import4803100%11,440Massive. 4 hours of manual entry per account
5Mobile offline mode3,400280%6907High. Top reason technicians abandon the app
6QuickBooks sync1,650380%5792Massive. Replaces a manual monthly export
7Route optimization2,900350%9483Massive. One extra job per technician per day
8Custom report builder2,100250%8263High. Replaces ad hoc report requests

Every input carries its own basis column; only impact is shown here for width. The space rule blocks a score whose basis is empty, because a reach figure from the billing export and one somebody felt strongly about look identical in a total.

Sensitivity Check

For each item, the cheapest single change to one input that gains it a rank. Derived entirely from Scored Backlog.

RankRequestGapCheapest single changeNew scoreNew rankVerdict
1SMS remindersfirstImpact down one notch, medium to low2,0003rdFragile
2Two-factor auth1,800Impact up one notch, low to medium4,4001stFragile
3Photo attachments120Confidence up one notch, medium to high2,6002ndFragile
4Bulk import640Effort down one step, 1 month to half2,8802ndFragile
5Mobile offline533None. Impact to massive still lands at 1,360 and stays 5th. Effort must fall below 3.8 months, a 37% cut1,3605thFirm
6QuickBooks sync115Confidence up one notch, medium to high9905thFragile
7Route optimization309None in one notch. Confidence to medium reaches 773 and stays 7th. Low to high, two notches, reaches 9677737thFirm
8Custom reports221None in one notch. Impact to massive reaches 394 and stays 8th. Confidence two notches reaches 5253948thFirm

First and second place are separated by one notch of the most subjective input in the formula. That is the headline, and it is uncomfortable: rate two-factor medium instead of low and it is first outright at 4,400. Ranks five, seven and eight are the only positions here that can be defended as arithmetic.

Estimate Disagreements

Both values, both bases, and the rank each one produces. The gap between those ranks is what settling the argument is worth.

RequestInputValue AValue BRank ARank BWorthHow to settle it
QuickBooks syncReach1,6504,3006th4th2 placesSurvey had a 38% response rate. Ask the other 62% of accounts.
Route optimizationConfidence50%100%7th5th2 placesMeasure it. Two weeks of route data from 20 of our own technicians.
Two-factor authImpact0.512nd1st1 placeDecide whether impact means impact on the user or on the goal.
Mobile offlineEffort635th4th1 placeNot one estimate twice. Read-only offline and full sync are two rows.
Custom reportsImpact238th8th0 placesNothing to settle. Effort and confidence hold it last, so winning this changes nothing.

The last row is the most useful thing this sheet produces. Sales is arguing that reporting deserves a higher impact score, and they may be right, and it would not move the item, because eight person-months against unspecified reports is what holds it last. Telling them that redirects the conversation to the input that would.

What you get

01

Sensitivity Check

For every item, the smallest single change to one input that would gain it a rank, the score that change produces, and a verdict of Fragile, Contested or Firm. Every figure recomputes from the scored inputs, so a challenger argues with the method rather than with you.

02

Scored Backlog

The four inputs with a basis column beside each one. A reach figure from the billing export and a reach figure somebody felt strongly about look identical in a total, which is why the space rule blocks a score whose basis is empty.

03

Estimate Disagreements

Two values for one input, side by side with both bases, and the rank each one produces. The gap between those ranks is exactly what settling the argument is worth, which is how you decide whether to have it.

04

Request Intake

Every request in the words it arrived in, with its channel and evidence. Requests that cannot be scored stay here and say why, rather than getting an invented reach figure so they can sort alongside the real ones. Where the raw material is a support queue rather than a list, feedback triage names the requests first, and voice of customer reconciles channels that disagree.

05

Scoring Method

The definitions this backlog was actually scored against, including the one that matters most: whether impact means impact on the user or impact on the goal. Those two questions give different orders, and mixing them makes a list that is not an order at all.

06

Decision Log

What was decided, what rank it overrode, and who decided it. Working out of order is normal and often right. Doing it quietly is what makes next quarter's scores worthless. The order then feeds the roadmap pack, which checks whether it fits.

How it works

  1. 1

    Settle what impact means

    River asks one question first, because it decides more about the final order than any individual score: impact on the user who meets the change, or impact on the company goal.

  2. 2

    Send the backlog

    A Productboard or Canny export, the spreadsheet of sales asks, support tags, meeting notes. River fills intake with what was actually said and flags duplicates, overlaps and anything not yet scorable.

  3. 3

    Get it scored with the evidence

    Every input gets a basis. Where no team effort estimate exists, River leaves it blank and lists the rows that need one rather than inventing the weakest number in the formula.

  4. 4

    Find out what is fragile

    River searches all four inputs on every item and reports how much of the order is judgement. Expect the top two to be close; that is the useful outcome, not a failure.

Frequently asked questions

Is this RICE, or something else?

It is RICE, with the scales as Intercom originally published them, plus the part they left out. Their post is explicit that scores "shouldn't be used as a hard and fast rule" and that working out of order is fine. This pack takes that seriously by measuring how far from a rule any given order actually is.

Why does impact cause so much of the trouble?

Because it is the only input on an invented scale and the scale gets multiplied. Massive is 3, high 2, medium 1, low 0.5, minimal 0.25. Three of those four steps double the score. So a single notch, chosen by one person in one afternoon, can outweigh a carefully measured reach figure.

Does fragile mean the score is wrong?

No. Most items in most backlogs come out fragile, because that is what multiplying an ordinal scale does. Fragile means the position cannot be defended as a calculation, so defend it as a judgement instead: name whose, on what basis. That is a stronger position than pretending to arithmetic you do not have.

Can it use a framework other than RICE?

Yes. Weighted scoring, value versus effort, cost of delay divided by duration, or your own house model. The sensitivity search is indifferent to which inputs exist, since it moves one at a time and recomputes. Tell River the model in the first session and Scoring Method records it.

Who provides the effort estimates?

Your engineering team, and this space deliberately does not produce them. An effort figure River invented would be both the weakest input in the formula and the one you would have to defend. Rows without an estimate stay unscored and get listed, which is more useful than a rank resting on a guess.

How is this different from a roadmap?

An order is not a plan. This space says what should come first and why. Whether the team can actually deliver the top six, and what each line commits you to, is the roadmap pack. Specifying one of them in enough detail to build is the PRD pack. They stack in that order.

What if leadership just overrides the list?

Then record it, which is what Decision Log is for. The 2020 Scrum Guide makes ordering the Product Owner's accountability rather than a formula's, so an override is legitimate by design. What is not legitimate is reordering quietly and leaving the scores in place, because after two rounds nobody believes the sheet. Knowing which override is coming is preparation rather than scoring.

Find out how much of your order is real

Send the backlog in whatever shape it lives in. River scores it with the evidence attached, then tells you which single estimate would reorder the list.

Score my backlog