RICE scoring ranks features by Reach x Impact x Confidence / Effort, giving each a comparable priority number. It works because it forces you to estimate value and cost explicitly instead of arguing from opinion — but the output is only as honest as your inputs. If Reach and Impact are guesses pulled from a gut feeling, RICE just launders bias into a number that looks objective. Pair it with real usage and revenue data and it becomes one of the more reliable ways to rank a backlog.
What does RICE actually stand for?
RICE is an acronym for four factors, each scored independently, then combined into a single priority score:
- Reach — how many customers or accounts this will affect in a given time period (e.g., per quarter).
- Impact — how much it moves the needle for each person reached, when it lands.
- Confidence — how sure you are about your Reach and Impact estimates.
- Effort — the total person-time to build and ship it, usually in person-months.
The model was popularized by Intercom's product team as a way to compare wildly different feature requests on the same scale — see the Intercom blog for the original framing.
How do you calculate a RICE score?
The formula is:
RICE Score = (Reach × Impact × Confidence) / Effort
Here's how each factor is typically scored:
| Factor | Scale | What it measures |
|---|---|---|
| Reach | Number of users/accounts per quarter (e.g., 50, 500, 5000) | Breadth of impact |
| Impact | 3 = massive, 2 = high, 1 = medium, 0.5 = low, 0.25 = minimal | Depth of impact per user |
| Confidence | 100% = high, 80% = medium, 50% = low | How solid your evidence is |
| Effort | Person-months (e.g., 0.5, 1, 3, 6) | Cost to build |
Worked example
Say you're comparing two features:
- Bulk export (Feature A): Reach = 800 accounts/quarter, Impact = 1 (medium), Confidence = 80%, Effort = 2 person-months.
Score = (800 × 1 × 0.8) / 2 = 320 - Custom SSO (Feature B): Reach = 40 accounts/quarter, Impact = 3 (massive), Confidence = 100%, Effort = 1.5 person-months.
Score = (40 × 3 × 1) / 1.5 = 80
Bulk export wins numerically, even though SSO feels more strategic to the six enterprise accounts asking for it. This is exactly where RICE needs a second lens — more on that below, and in our guide on how to weight requests by revenue.
Where can I get a RICE prioritization template?
A usable RICE template needs five columns per feature: Reach, Impact, Confidence, Effort, and the calculated Score, plus a notes column for evidence sources (support tickets, NPS verbatims, sales-loss reasons). Structure it as:
- List every candidate feature as a row.
- Fill in Reach using actual usage/account data, not estimates from memory.
- Score Impact on the 0.25–3 scale, and write down why.
- Set Confidence based on how much evidence backs Reach and Impact — cite the source.
- Estimate Effort with your eng lead, in person-months.
- Let the sheet auto-calculate the score and sort descending.
Build this as a shared spreadsheet so scoring assumptions are visible and challengeable — a static ranked list without visible inputs just becomes another opinion. Our prioritization matrix template extends this into a full comparison view across multiple frameworks.
What's a good RICE score?
There's no universal "good" number — RICE scores are only meaningful relative to other items in the same backlog, scored with the same units and time window. A score of 320 is high if your other candidates cluster around 50, and mediocre if your top feature scores 1,200. What matters more than the absolute number is consistency: use the same Reach time window (e.g., always "per quarter") and the same Impact scale across every item, or the comparison breaks. Re-run RICE scoring every planning cycle rather than treating scores as permanent — Reach and Effort estimates get more accurate as you build.
What are RICE's limitations, and where does it break?
RICE's biggest weakness is confidence gaming: because Confidence is self-reported and multiplicative, a team that wants a pet feature prioritized can simply mark Confidence at 100% with no real evidence behind it, inflating the score to whatever they need. There's no built-in check against this unless someone audits the inputs.
Other common failure modes:
- Effort is undercounted. Person-months rarely include QA, docs, support ramp-up, or the maintenance tax — so Effort is systematically too low, inflating every score.
- Reach conflates volume with value. 1,000 free-tier users asking for dark mode isn't the same as 20 enterprise accounts asking for SAML, but raw Reach treats them identically.
- No revenue dimension. Vanilla RICE ignores deal size, expansion potential, and churn risk entirely — it optimizes for "affects the most people," not "protects or grows the most revenue."
- Impact is subjective by design. The 3/2/1/0.5/0.25 scale is a judgment call dressed up as a number, so two PMs scoring the same feature can land far apart.
The fix for most of these is to add a revenue-weighting layer on top: multiply Reach by average account value (or replace it with ARR touched) so the score reflects dollars, not just headcount. This is the approach we cover in weighting requests by revenue, and it's also how you connect prioritization scores to actual prioritization ROI — the metric leadership actually cares about.
How is RICE different from ICE?
ICE (Impact, Confidence, Ease) is RICE's simpler, faster cousin — it drops Reach entirely and swaps Effort for "Ease" (scored 1–10, higher is easier). ICE is quicker to run because it skips the reach-estimation step, which makes it useful for rapid backlog triage or early-stage teams without much usage data. RICE takes longer to calculate but produces more defensible scores because Reach forces you to quantify "how many people does this actually touch" instead of relying on a single ease/effort gut check. If you're prioritizing dozens of small requests quickly, use ICE. If you're making a quarterly roadmap bet that leadership will scrutinize, use RICE — and ideally with revenue weighting layered in. Frameworks like RICE and ICE both get more rigorous treatment in resources like ProductPlan, Reforge, and SVPG, which is worth reading if you want to go deeper on prioritization theory beyond a single scoring formula.
RICE isn't magic — it's a forcing function. The value isn't the formula itself, it's the discipline of writing down your Reach, Impact, and Effort assumptions where someone else can challenge them. Combine it with real usage data, a revenue weighting layer, and a healthy skepticism toward 100% Confidence scores, and it becomes one of the more defensible ways to justify what you build next. For a broader look at collecting and triaging the requests that feed into RICE in the first place, see our guide on how to prioritize feature requests.
See how VocxAI builds this for you
VocxAI connects your customer signals to your revenue data and surfaces a ranked, revenue-weighted product backlog - automatically, every week.
Join the private beta