Instrument Metrics

Instrument Metrics

When to use

Map each KR to a measurement spec precise enough two people calculate the same number — with cross-plugin handoff to performance when installed.

What this skill does not do

• Does not duplicate performance:metrics-glossary when performance plugin installed — mapping only.
• Does not set targets — route to /okr:set-targets after baselines known.
• Does not invent data sources — flag unknowns for user input.

Preconditions

| Input | If missing | |---|---| | KRs to instrument | Ask user to provide | | Practice profile (performance plugin status, check-in frequency) | Detect performance plugin; flag [PROVISIONAL] on cadence match |

Provisional mode

Ambiguous formulas: flag AMBIGUITY FLAGS — do not pick calculation method silently.

Trust spine

• Confidence bands (structured-aggregation):
• High: Every KR mapped; formula/source/owner/cadence complete or deferred to performance with explicit handoff.
• Medium: Some ambiguities flagged; cadence mismatches noted.
• Low: KRs without measurable spec — halt target-setting.
• Failure modes:
• Strategic advice vs. support: Spec is draft for data owner validation.
• Client confidentiality: Metric definitions may expose internal systems — CONFIDENTIAL header.
• Accountability gap: Owner named per metric; ambiguities flagged not hidden.
• Analytical Rigor: Completeness check — every KR has mapping or explicit gap.
• Incentive Gaming: N/A — definition focus.
• Escalation triggers: Check-in frequency faster than metric refresh — flag mismatch blocking reliable pulse.

Workflow

1. Read practice profile for performance plugin status.
2. If performance installed: KR-to-metric mapping only; defer definition to performance:metrics-glossary.
3. If performance not installed or metric new: full spec — formula, source, owner, refresh, caveats.
4. Cadence match check — check-in frequency vs. metric refresh.
5. Flag ambiguous formulas — do not pick silently.
6. Completeness/source-tag check before output.

Output format

CONFIDENCE: [defensible recommendation | structured first pass]
KR: [text]

[If performance installed:]
  Mapped metric: [name] — full definition owned by performance:metrics-glossary
  Cadence match: [ok | mismatch — explain]

[If performance not installed, or metric is new:]
  Formula: [exact calculation]
  Source: [system/sheet/table]
  Owner: [name/role]
  Refresh: [cadence]
  Cadence match: [as above]
  Caveats: [...]

[repeat per KR]

AMBIGUITY FLAGS: [any KR where metric admits more than one calculation method]

Worked example

Input: KR "Enterprise NPS ≥ 50." Performance plugin installed. Weekly check-ins. NPS refreshes monthly in CRM.

Expected output (excerpt):

KR: Enterprise NPS ≥ 50
  Mapped metric: Enterprise NPS — full definition owned by performance:metrics-glossary
  Cadence match: mismatch — weekly check-in vs. monthly refresh [review]

AMBIGUITY FLAGS: none

Quality checks before delivering

• [ ] Performance handoff respected when plugin installed
• [ ] Full spec when performance not installed
• [ ] Cadence match checked per KR
• [ ] Ambiguities flagged, not silently resolved
• [ ] Every KR covered

Propose profile update

When a stable convention surfaces during this run (thresholds, naming, tone, output format, or recurring corrections), propose a profile update: show the exact diff against ~/.claude/plugins/config/claude-for-strategy/okr/CLAUDE.md (org-wide facts go to org-profile.md), ask for confirmation, and write only on yes. Only /okr:practice-setup auto-applies a full profile write.

Outputs

Follows plugin CLAUDE.md § Outputs. Next: performance:metrics-glossary, set-targets, or resolve cadence mismatch.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: daddia
• Source: daddia/claude-for-strategy
• License: MIT

Install and usage instructions live in the source repository linked above.