What the number on the slider means

Updated October 23, 2026 to describe the current scoring model and correct earlier claims about comparability.

A reach score of 55 does not mean 55 people. An effort score of 35 does not mean 35 days. Without a shared definition, those numbers leave a product discussion with more interpretation than it needs.

What the scale represents

Tarazu uses a band-based adaptation of RICE. Every input sits on a 1–100 scale, and the app computes (Reach × Impact × Confidence) ÷ Effort, rounded. It multiplies the band scores, rather than raw audience counts, outcome changes and person-days.

Putting different inputs on the same numeric range does not establish that this formula preserves classic RICE ratios, removes unit mismatch, or makes judgments comparable across teams and quarters. An earlier version of this post overstated those benefits. The resulting index is a heuristic for comparing supplied judgments. A score twice as large does not mean twice the benefit.

Give each number a basis

Set a shared audience, scoring period and target outcome. Then record a candidate's quantities, scope, evidence and uncertainties in claims. A score can point to one existing claim as its recorded basis; the app shows its words and provenance beside the current band definition.

For example, a fictional forecast of 60 of 100 users over 30 days can explain Reach 55, the Broad band. An estimate of three total person-days can explain Effort 35, the S band under the default rubric. The quantities remain forecasts or estimates. Selecting a claim neither verifies it nor changes the score automatically.

These defaults give the discussion a starting vocabulary:

• Reach: Niche, Segment, Broad, Majority, Everyone, with audience-share definitions.
• Impact: Minimal, Low, Medium, High, Massive; workspace definitions can specify the target outcome change.
• Confidence: Assumed, Directional, Supported, Measured. These describe evidence strength, not success probabilities.
• Effort: XS through XL, with total person-day ranges across roles. Workspace definitions can change those ranges.

Keep uncertainty visible

AI score proposals use listed anchors and should ask targeted questions when missing context could change a band. Manual values can still sit between anchors; the app labels them custom scores and does not round them into a named band.

Read the full selected claim, its linked sources and conflicting evidence before trusting a number. A reported-fact label is a user assertion, and a matching excerpt alone does not prove the claim. Changing the rubric or delivery scope calls for reviewing the comparison again.

The RICE scoring guide explains the formula, its limits and the difference from classic RICE. The purpose of the scale is to make a judgment discussable and revisitable, with enough context to challenge it.