Synthetic worked example · Not customer data
A complete record. A partial result.
Harborline Studio pilot review · Meeting: 4 Mar 2026
Fictional meeting owner and decider: Mara Chen.
Decision evidence
“Mara will own a two-week onboarding pilot with six clients. We’ll review activation on 18 March; Ellis still questions whether the sample represents larger clients.”
- Decision
- Mara will own a two-week onboarding pilot with six clients.
- Unresolved risk
- Small clients may not represent larger accounts.
- Actions with owners
- Recruit clients — Mara Chen
Collect activation data — Ellis Rao - Extracted signals
- Confidence: medium
Consensus: unclear
Check-in due 18 Mar 2026; outcome recorded 20 Mar 2026 (late). Coverage snapshot: 21 Mar 2026.
Outcomes 1/1 · Due check-ins 1/1
Average of four 100% components.
1 decision, 1 gap: unclear consensus.
100 − int((1/5) × 80) = 84.
One risk alone adds no risk penalty.
Only four of six clients activated. The pilot missed its target, and the representativeness risk remains unresolved. Complete coverage is not business success.
What do these scores mean?
Meeting Governance measures recorded coverage: action items with a nonblank owner, decisions with a decider, decisions with an outcome record, and due check-ins with an outcome record. Each percentage is rounded; the score is the rounded average of components with a nonempty denominator. Empty components are omitted, not treated as zero.
Due check-in coverage counts decisions whose check-in date is today or earlier and that currently have an outcome record. It does not compare the outcome timestamp with the deadline. A late record counts; this is not evidence of punctuality or success.
Decision Intelligence (DI) health is a different rule-based gap score: 100 - int(min(gaps / (5 * decisions), 1) * 80). Here int truncates the penalty. Checks look at evidence, consensus, confidence, ownership wording in decision text, and multiple unresolved risks. Structured action-owner coverage is not the DI ownership-wording check.
No decisions? A DI report has no numeric health score. Governance can be computed from action ownership alone, but the meeting header only displays it when an intelligence report exists. With no decisions or actions, governance has no score either.
Confidence is separate. The extracted confidence signal describes apparent commitment in the meeting, not either score, a verified fact, or the probability of a successful business outcome. People must review the source evidence, owners, dates, risks, and actual outcomes. High scores can coexist with failure.
What if the extraction is wrong?
Supported action-owner correction: with meeting edit access, open the meeting → Action Items tab → Edit on “Collect activation data” → replace the Owner field with “Ellis Rao” → Save. This edits the action’s owner text, not the decision’s decider or DI ownership wording.
Decision correction is different: there is no decision-text editor or arbitrary decider picker in the current UI. The Pro check-in API is restricted to the decision owner and permits only self-assignment as decider. It cannot assign Ellis as Mara’s decider. Verify incorrect decision text or another person’s authority against the source and keep a separate correction record; do not treat this sample as an editing workflow for unsupported fields.
