Impact
Expected effect on one agreed target outcome, activation, conversion, retention, or revenue. Impact on what? Without a named outcome, two scorers can enter the same number for entirely different reasons.
ICE Score Calculator (Impact, Confidence, Ease)
This ICE score calculator uses Impact Confidence Ease for feature prioritization in your product backlog, with one difference from most calculators: it keeps ICE fast while making the scoring contract and human disagreement inspectable.
ICE is valuable because it makes a rough judgment explicit and fast. Its weakness is not lack of decimals; it is hidden differences in what people mean by Impact, Confidence, and Ease. Define the contract, score independently when it matters, expose disagreement instead of averaging it away, and declare your formula variant.
- Scoring contract: shared Impact, Confidence, and Ease anchors plus one target outcome
- Independent team scoring with reveal, visible disagreement, and explicit reconciliation
- Average (default) and Multiply variants with an order-based variant audit, scores never compared across variants
- Import, export, and shareable ranking workflows with local-only storage
No login. Inputs stay in this browser and are autosaved locally; CraftUp does not persist initiative text on its backend.
Formula variant
Average (default): (Impact + Confidence + Ease) / 3. Output scale 1-10 on 1-10 inputs. Both current methodology sources that state a formula use the average; the primary source supports the dimensions, not the arithmetic.
Mode
Ease mode scores delivery simplicity directly on a 1-10 scale defined by your scoring contract.
Usage Mode
0 assessed, 1 unassessed in this comparison.
Scoring contract, define what the numbers mean before rankingNot defined yet
ICE is a relative heuristic inside a defined scoring contract. A 7 from one person means something completely different from a 7 from another person unless the team writes down shared anchors first. These definitions are stored with your workspace and exported with the review, never sent to analytics.
Confidence here is a heuristic judgment about evidence quality, never statistical confidence, A/B significance, or a probability of success. Define the target outcome and all nine anchors to complete the contract.
Batch table with relative ranking
New rows start unassessed, a blank cell means “not judged yet”, never zero. Expand a row for rationale, candidate type, hard-constraint flag, and independent team scoring.
| Initiative | Impact | Confidence | Ease | ICE (Average) | Rank | Detail | Actions |
|---|---|---|---|---|---|---|---|
Unassessed | Ease: - | - | - |
Ranking uses unrounded arithmetic; rounding is display-only. Equal raw scores share a rank and are labeled as ties, input order never implies priority.
Higher-Ease candidates in this comparison
Descriptive and relative to this set only, no universal quick-win cutoff.
- No assessed candidates yet.
High-Impact / lower-Confidence combinations
Worth an explicit next action, experiment, prototype, research, instrumentation, defer, or reject. Never an automatic verdict.
- No assessed candidates yet.
Next decision, hand off the broader trade-offs
Define the scoring contract first. When the relative order is set, broader constraints still belong elsewhere. ICE gives a useful relative heuristic; the Product decision remains yours.
Preview Markdown decision review
Loading saved data...
How it works
Step 1
Define the scoring contract: one target outcome plus what low, mid, and high mean for Impact, Confidence, and Ease in this decision.
Step 2
Score candidates, independently before reveal when the decision matters, with one sentence of rationale per factor.
Step 3
Reveal disagreement, reconcile explicitly agreed values, and confirm one formula variant for the cycle.
Step 4
Inspect the relative order and the formula-variant audit, then hand broader trade-offs to prioritization or experiment design.
What ICE stands for
Confidence
A heuristic judgment about how much evidence supports the Impact and Ease assumptions. It is not statistical confidence, not A/B significance, and not a probability of success. If you need experiment statistics, use the A/B test plan generator.
Ease
Delivery simplicity as your team defines it, time, complexity, coordination, dependencies. There is no universal duration scale; write down what Ease means for this decision before scoring.
Why calibration matters more than arithmetic
A perfectly implemented formula still produces nonsense when one person scores Ease as elapsed days and another scores it as engineering complexity, or when Confidence means “I like it” to one scorer and “we have direct data” to another. A fast score without shared anchors creates numerical consistency without semantic consistency. Semantic comparability comes before arithmetic, always.
Formula variants, declared, not hidden
Sources disagree on ICE arithmetic: some describe the average, others multiply, and the primary source supports the three dimensions without establishing one canonical formula. This tool treats that ambiguity as real methodology landscape, not as a detail to hide.
Average (default)
ICE = (Impact + Confidence + Ease) / 3. Output scale 1–10 on 1–10 inputs. Default because the methodology sources that state a formula explicitly describe the average.
Multiply (alternate)
ICE = Impact × Confidence × Ease. Output scale 1–1000 on 1–10 inputs. Neither variant is universally correct, pick one per review cycle and never compare raw scores across variants, because they do not share a scale.
The in-tool variant audit recomputes your reconciled inputs under the other formula and compares the ordering, not magnitudes. If the top order changes, the variant choice was decision-relevant, and now you can see it.
Worked examples (all hypothetical)
Same numbers, different meaning
Team A writes “Ease 8 = testable in under one week with one team.” Team B writes “Ease 8 = no cross-team dependency, regardless of duration.” Same number, different meaning. Lesson: compare rows only inside one explicit scoring contract.
Scorer disagreement is information
Simplifying an onboarding step: Scorer A enters Impact 8, Confidence 7, Ease 8; Scorer B enters Impact 4, Confidence 6, Ease 8. The gap comes from different assumptions, A believes the step is the main activation blocker, B saw analytics showing drop-off happens later. Lesson: expose the 8-vs-4 Impact disagreement and reconcile the assumption before reconciling the number.
Formula choice can move the order
A spiky 10/10/1 candidate outranks a steady 6/6/6 candidate under Average (7.0 vs 6.0), but the order flips under Multiply (100 vs 216). Neither order is the correct Product answer. Lesson: declare one variant deliberately and check the audit before presenting a rank as stable.
Hard constraints stay outside the score
A high-ICE growth experiment and a mandatory security remediation sit in the same table. The remediation may “win” on numbers, and it does not matter. Lesson: never manufacture scores to make a mandatory item win; flag it as a hard constraint and keep it outside the discretionary comparison.
Loadable examples
Simplify onboarding step (scorer disagreement)
Hypothetical: reconciled Impact 8, Confidence 7, Ease 8, but only after two scorers disagreed 8 vs 4 on Impact and discussed the activation-bottleneck assumption.
A relative rank under your scoring contract, not a verdict. The value of this example is the disagreement it surfaced, not the number it produced.
Balanced vs spiky bet (formula-variant check)
Hypothetical: steady 6/6/6. Compare it against a spiky 10/10/1 candidate to see the Average vs Multiply order change in the formula-variant audit.
Order moves across variants while inputs stay identical. Pick one variant deliberately per review cycle; never compare raw scores across variants.
Mandatory security remediation (outside ICE)
Hypothetical: high ICE numbers on a mandatory item. Kept in the table only to demonstrate the hard-constraint flag, exclude it from the ranking.
Flagged as a hard constraint and excluded from the relative order. ICE ranks discretionary choices; obligations do not need a score to matter.
ICE vs RICE, and when not to use ICE
ICE's practical advantage is fast relative scoring without a separate Reach input. Use ICE when candidates are reasonably comparable, fast heuristic ranking is valuable, and Reach is not being modeled explicitly. Use RICE when Reach differs materially across candidates and you can estimate it meaningfully. Neither framework is always more rigorous, and neither replaces the other.
Do not force mandatory work through ICE, incidents, legal deadlines, security remediation, and contractual commitments belong outside the score. And a high relative ICE score is never a roadmap commitment: strategy, constraints, dependencies, and opportunity cost live in the Prioritization Workspace, which owns the broader Product choice.
Pro tips
- Define what Impact, Confidence, and Ease mean for this decision before scoring a single row.
- Anchor Impact to one agreed target outcome so scores stay comparable across candidates.
- Treat Confidence as a heuristic evidence judgment, never as statistical confidence.
- Score independently before discussing when the decision matters; anchoring is the most common failure mode.
- When scorers disagree, ask what assumption explains the difference instead of averaging it away.
- Pick one formula variant deliberately per review cycle and never compare raw scores across variants.
- Write one sentence of rationale per factor so future-you can audit the rank.
- Revisit scores when evidence or context changes materially, not on a fixed calendar.
- Keep candidate types comparable; split the set when experiments and platform bets are mixed.
- Hand strategy, constraints, and opportunity cost to broader prioritization, ICE only produces a relative order.
Common mistakes
Symptom: Two teammates both score Impact 8 but mean completely different things.
Cause: No shared scoring contract: one means conversion lift, the other means qualitative value.
Fix: Define low/mid/high anchors for each factor and the target outcome before ranking.
Symptom: The first score stated in the meeting becomes everyone's score.
Cause: Scores are entered publicly, so anchoring replaces independent judgment.
Fix: Enable independent team scoring, hide entries until reveal, then discuss the raw disagreement.
Symptom: Averaging two scorers produces a number nobody believes.
Cause: Mathematical averaging hides the assumption gap instead of resolving it.
Fix: Keep raw scorer values visible and explicitly reconcile one agreed value per factor.
Symptom: Rank #1 is treated as a roadmap commitment.
Cause: A relative heuristic is read as an objective decision.
Fix: Present ICE as a relative order under one contract; route strategy and constraints to broader prioritization.
Symptom: Switching formula changes the winner and nobody noticed.
Cause: Average and Multiply do not share a scale, but raw scores were compared across variants.
Fix: Use the formula-variant audit to compare ordering only, and declare one variant per cycle.
Symptom: Effort in days was converted into Ease automatically.
Cause: The 11-minus-effort convenience was applied to real effort units it was never designed for.
Fix: Use Effort mode only when your 1-10 Effort scale is the inverse of Ease; otherwise define a team Ease anchor.
Symptom: A mandatory compliance item keeps losing, so the team inflates its scores.
Cause: Hard constraints are forced through a discretionary ranking tool.
Fix: Flag hard constraints and keep them outside the ICE comparison instead of manufacturing a win.
Symptom: Import fails or rows look broken.
Cause: CSV headers are missing or values fall outside the 1-10 range.
Fix: Download the template, keep the headers, and leave cells blank for candidates you have not assessed yet.
FAQ
Is this ICE score calculator free?
Yes. This ICE score calculator is fully free, requires no login, and supports unlimited initiatives. You can define a scoring contract, score a single idea or rank a full batch with optional independent team scoring, then export the decision review without any paywall steps.
Do you store my data?
CraftUp does not persist your initiative text on its backend. Inputs stay in this browser and are autosaved locally so you can resume work, and you can clear them anytime. Share links encode the workspace in the URL itself, so review sensitive text before sharing externally.
How is this different from using ChatGPT directly?
Chat tools produce plausible prose but do not enforce a shared scoring contract or expose scorer disagreement. This page keeps one explicit Impact, Confidence, and Ease definition, supports independent scoring before reveal, audits formula-variant order changes, and exports a traceable review in one place.
Can I use this for work or client projects?
Yes. The outputs are designed for practical product planning, growth experiment ranking, and backlog reviews in client or internal environments. Record per-factor rationale and the scoring contract so stakeholders can audit the rank later. Export Markdown directly into tickets, docs, or stakeholder updates.
Which formula should I choose: Multiply or Average?
Average computes (Impact + Confidence + Ease) / 3 on a 1-10 scale, while Multiply computes Impact x Confidence x Ease on a 1-1000 scale. Neither is universally canonical, so pick one variant deliberately per review cycle and never compare raw scores across variants because they do not share a scale.
What is the difference between Ease mode and Effort mode?
Ease mode scores delivery simplicity directly from 1 to 10. Effort mode is a CraftUp convenience conversion on a deliberately inverse 1-10 scale, where Ease equals 11 minus Effort. Use it only when your Effort scale is actually defined as the inverse of Ease, never for real units like days or story points.
Can I import an existing backlog from a spreadsheet?
Yes. Use the CSV import option with the provided template headers. The tool accepts the legacy compact schema as well as the extended schema with candidate type, target outcome, and rationale fields, then recomputes ranking from raw precision. Blank cells stay unassessed instead of silently becoming default scores.
How should I score confidence realistically?
Treat Confidence as a heuristic judgment about how much evidence supports your Impact and Ease assumptions, never as statistical confidence or a probability of success. Define what low, mid, and high evidence mean for this decision in the scoring contract, and revisit the score when the evidence actually changes.
Does the share link include my notes?
Yes. The share URL encodes the workspace, rows, rationale, scoring contract, and scorer entries, with client-side compression in the URL itself. Opening the link reconstructs the same context in a fresh browser. Because anyone with the URL can reconstruct its contents, review internal text before sharing.
When should I choose ICE instead of RICE?
Use ICE when candidates are reasonably comparable, fast heuristic ranking is valuable, and Reach is not being modeled explicitly. Use RICE when Reach differs materially across candidates and you can estimate it meaningfully. For strategy, dependencies, and opportunity cost, use the broader Prioritization Workspace instead.
Learn more with CraftUp
Turn a Calibrated Rank Into a Decision
ICE gives you a relative order under one explicit contract. Take strategy, constraints, and opportunity cost to the Prioritization Workspace, or turn a testable candidate into an experiment plan.
Freshness
Last updated: 2026-09-22
- Added an explicit scoring contract: teams define what Impact, Confidence, and Ease mean before ranking.
- Added independent team scoring with reveal, visible disagreement, and explicit reconciliation, no auto-averaging.
- Made formula variants transparent with a default Average, an order-based variant audit, and no cross-variant score comparison.
- Removed universal priority bands, quick-win and research cutoffs, and the arbitrary confidence -2 check; new rows start unassessed.