metric-definition

Metric Definition

What You'll Need

Input	Required?	Example
Metric name(s) to define	✅ Required	"Weekly Active Users", "Activation Rate"
Product / context	✅ Required	B2B SaaS project management tool
Data sources available	🟡 Recommended	PostgreSQL events table, Segment, Mixpanel
Team or squad that owns the metric	⚪ Optional	Growth team, Onboarding squad
Current tracking / calculation method	⚪ Optional	"We count DAU from our analytics dashboard"

Don't have everything? Start anyway — the skill will work with what you provide and flag where richer input would improve the output.

Purpose

Create precise, unambiguous metric definitions that a team can rally around. Eliminates the silent disagreements where two people say the same metric name but compute it differently.

Domain Context

Most metric confusion stems from one of three problems:

1. Vanity metrics — metrics that feel good but don't reflect real value (e.g., total registered users, page views without engagement context)
2. Ambiguous definitions — "Active user" means something different to every team member
3. Misaligned scope — tracking a metric nobody can influence or that doesn't connect to business outcomes

This skill uses a structured spec format that forces clarity on definition, formula, data source, and ownership before a metric gets into a dashboard.

Instructions

You are defining metrics for $ARGUMENTS.

Step 1: Classify Each Metric

For each metric provided, classify it:

Category	Definition	Examples
Operational / Actionable	Teams can directly influence this metric; it reflects real user behavior or business outcomes	Activation rate, feature adoption rate, trial-to-paid conversion
Vanity	Looks impressive but doesn't indicate health or progress; hard to act on	Total registered users, page views, social media followers
Leading indicator	Predicts future outcomes; moves before lagging metrics change	Users who complete onboarding → predicts retention
Lagging indicator	Reflects outcomes that already happened; slow to respond to changes	Annual revenue, annual churn rate
Health / guardrail	Monitors something you must not break; not a goal	Error rate, p99 latency, support ticket volume

Flag any vanity metrics and explain why they're vanity — then suggest an actionable alternative.

Step 2: Write the Metric Spec

For each metric, produce a complete spec:

## Metric: [Metric Name]

**Type**: [Operational / Vanity / Leading / Lagging / Health]
**Owner**: [Team or role responsible for moving this metric]
**Last updated**: [date]

### Definition
[One precise sentence. Avoid ambiguity. Define edge cases.]

Example of a good definition: "Percentage of new users who complete at least 3 core actions within their first 7 days after sign-up, where core actions are defined as: [action A], [action B], [action C]."

Example of a bad definition: "Users who are active."

### Formula
[Numerator] / [Denominator] × 100 (if percentage)
OR
[Precise calculation description]

**Numerator**: [precise definition]
**Denominator**: [precise definition]
**Excludes**: [what is excluded and why — e.g., internal users, test accounts]
**Time window**: [e.g., rolling 7 days, calendar month, event-based]

### Data Source
| Element | Source | Table / Collection | Field |
|---------|--------|-------------------|-------|
| [numerator element] | [database/platform] | [table] | [field] |
| [denominator element] | [database/platform] | [table] | [field] |

### Example SQL (if applicable)
```sql
-- [Metric name] calculation
SELECT
  [numerator query] AS numerator,
  [denominator query] AS denominator,
  ROUND(100.0 * [numerator] / NULLIF([denominator], 0), 2) AS metric_value
FROM [table]
WHERE [conditions]

Baseline & Targets

Period	Value	Notes
Current baseline	[value]	[context]
30-day target	[value]
90-day target	[value]

Review Cadence

• Tracked: [Daily / Weekly / Monthly]
• Reviewed in: [Weekly metrics review / Sprint review / Board meeting]
• Alert threshold: Trigger review if metric drops more than [X]% week-over-week

Related Metrics

• Drives: [North Star metric this feeds into]
• Driven by: [Input metrics that influence this]
• Counter-metric: [Metric that catches gaming — e.g., if activation rate = success metric, completion quality = counter]

Known Limitations

[What this metric doesn't capture. What could make it misleading.]

### Step 3: Flag Vanity Metrics

For any metric classified as vanity, explicitly note:
- Why it's vanity (what useful signal it lacks)
- The actionable alternative to track instead
- Whether to stop tracking it entirely or keep it as secondary context

### Step 4: Build the Metrics Dictionary (if multiple metrics)

If defining 3+ metrics, produce a summary table:

| Metric | Type | Owner | Formula (simplified) | Cadence | Dashboard |
|--------|------|-------|---------------------|---------|-----------|
| [Metric] | [Type] | [Owner] | [Formula] | [Freq.] | [Link] |

## Notes
- A metric without a formula isn't a metric — it's a concept
- Every metric should have exactly one owner; "everyone's" metric is nobody's metric
- Counter-metrics prevent Goodhart's Law: when a metric becomes a target, optimize the counter-metric too
- Revisit definitions quarterly — product changes often silently invalidate old formulas