Skip to content
Sparqbox Docs

How scoring math works

Sparqbox uses one formula for every score, every reviewer, every idea. The math is deterministic and the AI never calculates totals. This page walks the formula with a worked example.

The formula

Section titled “The formula”

For a single reviewer:

reviewer_total = SUM(score x weight) for each scored criterion

Where:

  • score is the reviewer's 1 to 5 answer per criterion.
  • weight is the decimal weight for that criterion, between 0 and 1.
  • Active criteria weights in a category sum to 1.000 (with a 0.001 rounding tolerance).
  • reviewer_total lands between 1.000 and 5.000 and is rounded to one decimal in the UI.

Worked example

Section titled “Worked example”

Three criteria groups in Process Improvement, six active criteria, one reviewer.

| Criterion | Weight | Score | Contribution | |---|---:|:---:|---:| | Efficiency Gain | 0.250 | 5 | 1.250 | | Implementation Effort | 0.150 | 3 | 0.450 | | Alignment | 0.200 | 4 | 0.800 | | Affected Scope | 0.150 | 4 | 0.600 | | Resource Availability | 0.150 | 3 | 0.450 | | Employee Benefit | 0.100 | 5 | 0.500 | | Total | 1.000 | | 4.05 |

The reviewer's weighted total is 4.05, displayed as 4.1. Because 4.1 >= 3.5 (the default approve threshold), the auto-decision rule fires approved.

Multiple reviewers

Section titled “Multiple reviewers”

When more than one reviewer scored the idea, the totals are combined.

  • Humans only, each human reviewer's total is computed, then averaged across humans, rounded to one decimal.
  • AI only, the AI's total is rounded to one decimal.
  • Humans and AI, a blend is used:
ai_w = workspace AI weight (default 1.00, range 0.00 to 1.00)
weighted = (human_avg x 1/(1 + ai_w)) + (ai_total x ai_w/(1 + ai_w))

With the default AI weight of 1.00, the AI counts equally with the human average. With 0.50, the AI counts half as much as the average human. With 0.00, the AI is excluded from the total.

The UI shows individual reviewer scores side by side and one combined weighted score used for the auto-decision.

The auto-decision rule

Section titled “The auto-decision rule”

Once the weighted score is computed, the system applies the category's thresholds.

  • weighted >= approve_threshold → approved
  • weighted <= reject_threshold → rejected
  • between the two → needs_discussion

Defaults are 3.5 for approve and 2.0 for reject. Admins can change both per category.

The auto-decision and the auto-feedback are written together; the submitter receives a decision email immediately.

Why weights, not straight averages

Section titled “Why weights, not straight averages”

Weights let you say what matters most for a given category. Approving a Cost Reduction idea on its strategic alignment alone might look generous; weighting Savings Potential at 0.250 and Payback Time at 0.150 makes the math reflect that. Straight averages flatten that signal.

Snapshots keep history stable

Section titled “Snapshots keep history stable”

Two safeguards prevent retroactive admin edits from rewriting old scores.

  • Per-idea snapshot, the set of criteria active at submission time is frozen onto the idea.
  • Per-row snapshot, every score row carries the weight that was in force when the score was entered.

So when an admin tunes weights or adds a criterion this week, last week's decisions don't change.

What's next

Section titled “What's next”