Agent-Skills.md

Agent Skills: OKR Design & Metrics Framework

OKR trees, KPI dashboards, North Star Metric, leading/lagging indicators, and experiment design. Use when setting team goals, defining success metrics, building measurement frameworks, or designing A/B experiment guardrails.

document-asset-creationID: yonatangross/orchestkit/okr-design

Repository

yonatangross/orchestkit
yonatangrossLicense: MIT
13314

Install this agent skill to your local

pnpm dlx add-skill https://github.com/yonatangross/orchestkit/tree/HEAD/src/skills/okr-design

Skill Files

Browse the full folder contents for okr-design.

Download Skill

Loading file tree…

src/skills/okr-design/SKILL.md

Skill Metadata

Name
okr-design
Description
"OKR trees, KPI dashboards, North Star Metric, leading/lagging indicators, and experiment design. Use when setting team goals, defining success metrics, building measurement frameworks, or designing A/B experiment guardrails."

OKR Design & Metrics Framework

Structure goals, decompose metrics into KPI trees, identify leading indicators, and design rigorous experiments.

OKR Structure

Objectives are qualitative and inspiring. Key Results are quantitative and outcome-focused — never a list of outputs.

Objective: Qualitative, inspiring goal (70% achievable stretch)
+-- Key Result 1: [Verb] [metric] from [baseline] to [target]
+-- Key Result 2: [Verb] [metric] from [baseline] to [target]
+-- Key Result 3: [Verb] [metric] from [baseline] to [target]

## Q1 OKRs

### Objective: Become the go-to platform for enterprise teams

Key Results:
- KR1: Increase enterprise NPS from 32 to 50
- KR2: Reduce time-to-value from 14 days to 3 days
- KR3: Achieve 95% feature adoption in first 30 days of onboarding
- KR4: Win 5 competitive displacements from [Competitor]

OKR Quality Checks

| Check | Objective | Key Result | |-------|-----------|------------| | Has a number | NO | YES | | Inspiring / energizing | YES | not required | | Outcome-focused (not "ship X features") | YES | YES | | 70% achievable (stretch, not sandbagged) | YES | YES | | Aligned to higher-level goal | YES | YES |

See references/okr-workshop-guide.md for a full facilitation agenda (3-4 hours, dot voting, finalization template). See rules/metrics-okr.md for pitfalls and alignment cascade patterns.

KPI Tree & North Star

Decompose the top-level metric into components with clear cause-effect relationships.

Revenue (Lagging — root)
├── New Revenue = Leads × Conv Rate          (Leading)
├── Expansion   = Users × Upsell Rate        (Leading)
└── Retained    = Existing × (1 - Churn)     (Lagging)

North Star + Input Metrics Template

## Metrics Framework

North Star: [One metric that captures core value — e.g., Weekly Active Teams]

Input Metrics (leading, actionable by teams):
1. New signups — acquisition
2. Onboarding completion rate — activation
3. Features used per user/week — engagement
4. Invite rate — virality
5. Upgrade rate — monetization

Lagging Validation (confirm inputs translate to value):
- Revenue growth
- Net retention rate
- Customer lifetime value

North Star Selection by Business Type

| Business | North Star Example | Why | |----------|--------------------|-----| | SaaS | Weekly Active Users | Indicates ongoing value delivery | | Marketplace | Gross Merchandise Value | Captures both buyer and seller sides | | Media | Time spent | Engagement signals content value | | E-commerce | Purchase frequency | Repeat = satisfaction |

See rules/metrics-kpi-trees.md for the full revenue and product health KPI tree examples.

Leading vs Lagging Indicators

Every lagging metric you want to improve needs 2-3 leading predictors.

## Metric Pairs

Lagging: Customer Churn Rate
Leading:
  1. Product usage frequency (weekly)
  2. Support ticket severity (daily)
  3. NPS score trend (monthly)

Lagging: Revenue Growth
Leading:
  1. Pipeline value (weekly)
  2. Demo-to-trial conversion (weekly)
  3. Feature adoption rate (weekly)

| Indicator | Review Cadence | Action Timeline | |-----------|----------------|-----------------| | Leading | Daily / Weekly | Immediate course correction | | Lagging | Monthly / Quarterly | Strategic adjustments |

See rules/metrics-leading-lagging.md for a balanced dashboard template.

Metric Instrumentation

Every metric needs a formal definition before instrumentation.

## Metric: Feature Adoption Rate

Definition: % of active users who used [feature] at least once in their first 30 days.
Formula: (Users who triggered feature_activated in first 30 days) / (Users who signed up)
Data Source: Analytics — feature_activated event
Segments: By plan tier, by signup cohort
Calculation: Daily
Review: Weekly

Events:
  user_signed_up  { user_id, plan_tier, signup_source }
  feature_activated { user_id, feature_name, activation_method }

Event naming: object_action in snake_case — user_signed_up, feature_activated, subscription_upgraded.

See rules/metrics-instrumentation.md for the full metric definition template, alerting thresholds, and dashboard design principles.

Experiment Design

Every experiment must define guardrail metrics before launch. Guardrails prevent shipping a "win" that causes hidden damage.

## Experiment: [Name]

### Hypothesis
If we [change], then [primary metric] will [direction] by [amount]
because [reasoning based on evidence].

### Metrics
- Primary: [The metric you are trying to move]
- Secondary: [Supporting context metrics]
- Guardrails: [Metrics that MUST NOT degrade — define thresholds]

### Design
- Type: A/B test | multivariate | feature flag rollout
- Sample size: [N per variant — calculated for statistical power]
- Duration: [Minimum weeks to reach significance]

### Rollout Plan
1. 10% — 1 week canary, monitor guardrails daily
2. 50% — 2 weeks, confirm statistical significance
3. 100% — full rollout with continued monitoring

### Kill Criteria
Any guardrail degrades > [threshold]% relative to baseline.

Pre-Launch Checklist

  • [ ] Hypothesis documented with expected effect size
  • [ ] Primary, secondary, and guardrail metrics defined
  • [ ] Sample size calculated for minimum detectable effect
  • [ ] Dashboard or alerts configured for guardrail metrics
  • [ ] Staged rollout plan with kill criteria at each stage
  • [ ] Rollback procedure documented

See rules/metrics-experiment-design.md for guardrail thresholds, performance and business guardrail tables, and alert SLAs.

Common Pitfalls

| Pitfall | Mitigation | |---------|------------| | KRs are outputs ("ship 5 features") | Rewrite as outcomes ("increase conversion by 20%") | | Tracking only lagging indicators | Pair every lagging metric with 2-3 leading predictors | | No baseline before setting targets | Instrument and measure for 2 weeks before setting OKRs | | Launching experiments without guardrails | Define guardrails before any code is shipped | | Too many OKRs (>5 per team) | Limit to 3-5 objectives, 3-5 KRs each | | Metrics without owners | Every metric needs a team owner |

Related Skills

  • prioritization — RICE, WSJF, ICE, MoSCoW scoring; OKRs define which KPIs drive RICE impact
  • product-frameworks — Full PM toolkit: value prop, competitive analysis, user research, business case
  • product-analytics — Instrument and query the metrics defined in OKR trees
  • write-prd — Embed success metrics and experiment hypotheses into product requirements
  • market-sizing — TAM/SAM/SOM that anchors North Star Metric targets
  • competitive-analysis — Competitor benchmarks that inform KR targets

Version: 1.0.0