Show which smaller drivers move a big product or business metric. — Claude Skill

---

name: metrics-tree description: Decomposes high-level North Star metrics into actionable sub-metrics and leading indicators, maps causal relationships between metric levels, and identifies high-impact experiments to move key metrics. Use when setting product North Star metrics, decomposing business metrics into drivers, mapping strategy to measurable outcomes, identifying which metrics to move through experimentation, understanding leading vs lagging indicators, prioritizing metric improvement opportunities, or when user mentions metric tree, metric decomposition, North Star metric, KPI breakdown, metric drivers, or how metrics connect.

Metrics Tree

Workflow

Copy this checklist and track your progress:

Metrics Tree Progress:
- [ ] Step 1: Define North Star metric
- [ ] Step 2: Identify input metrics (L2)
- [ ] Step 3: Map action metrics (L3)
- [ ] Step 4: Select leading indicators
- [ ] Step 5: Prioritize and experiment
- [ ] Step 6: Validate and refine

Step 1: Define North Star metric

Ask user for context if not provided:

• Product/business: What are we measuring?
• Current metrics: Any existing key metrics?
• Goals: What does success look like?

Choose North Star using criteria:

• Captures value delivered to customers
• Reflects business model (how you make money)
• Measurable and trackable
• Actionable (teams can influence it)
• Not a vanity metric

See Common Patterns for North Star examples by type.

Step 2: Identify input metrics (L2)

Decompose North Star into 3-5 direct drivers:

• What directly causes North Star to increase?
• Use addition or multiplication decomposition
• Ensure components are mutually exclusive where possible
• Each input should be controllable by a team

See resources/template.md for decomposition frameworks.

Step 3: Map action metrics (L3)

For each input metric, identify specific user behaviors:

• What actions drive this input?
• Focus on measurable, observable behaviors
• Limit to 3-5 actions per input
• Actions should be within user control

If complex, see resources/methodology.md for multi-level hierarchies.

Step 4: Select leading indicators

Identify early signals that predict North Star movement:

• Which metrics change before North Star changes?
• Look for early-funnel behaviors (onboarding, activation)
• Find patterns in high-retention cohorts
• Test correlation with future North Star values

Step 5: Prioritize and experiment

Rank opportunities by:

• Impact: How much will moving this metric affect North Star?
• Confidence: How certain are we about the relationship?
• Ease: How hard is it to move this metric?

Select 1-3 experiments to test highest-priority hypotheses.

See resources/evaluators/rubric_metrics_tree.json for quality criteria.

Step 6: Validate and refine

Verify metric relationships:

• Check correlation strength between metrics
• Validate causal direction (does A cause B or vice versa?)
• Test leading indicator timing (how early does it predict?)
• Refine based on data and experiments

Common Patterns

North Star Metrics by Business Model:

Subscription/SaaS:

• Monthly Recurring Revenue (MRR)
• Weekly Active Users (WAU)
• Net Revenue Retention (NRR)
• Paid user growth

Marketplace:

• Gross Merchandise Value (GMV)
• Successful transactions
• Completed bookings
• Platform take rate × volume

E-commerce:

• Revenue per visitor
• Order frequency × AOV
• Customer lifetime value (LTV)

Social/Content:

• Time spent on platform
• Content created/consumed
• Engaged users (not just active)
• Network density

Decomposition Patterns:

Additive Decomposition:

North Star = Component A + Component B + Component C

Example: WAU = New Users + Retained Users + Resurrected Users

• Use when: Components are independent segments
• Benefit: Teams can own individual components

Multiplicative Decomposition:

North Star = Factor A × Factor B × Factor C

Example: Revenue = Users × Conversion Rate × Average Order Value

• Use when: Components multiply together
• Benefit: Shows leverage points clearly

Funnel Decomposition:

North Star = Step 1 → Step 2 → Step 3 → Final Conversion

Example: Paid Users = Signups × Activation × Trial Start × Trial Convert

• Use when: Sequential conversion process
• Benefit: Identifies bottlenecks

Cohort Decomposition:

North Star = Σ (Cohort Size × Retention Rate) across all cohorts

Example: MAU = Sum of retained users from each signup cohort

• Use when: Retention is key driver
• Benefit: Separates acquisition from retention

Guardrails

Avoid Vanity Metrics:

• ❌ Total registered users (doesn't reflect value)
• ❌ Page views (doesn't indicate engagement)
• ❌ App downloads (doesn't mean active usage)
• ✓ Active users, engagement time, completed transactions

Ensure Causal Clarity:

• Don't confuse correlation with causation
• Test whether A causes B or B causes A
• Consider confounding variables
• Validate relationships with experiments

Limit Tree Depth:

• Keep to 3-4 levels max (North Star → L2 → L3 → L4)
• Too deep = analysis paralysis
• Too shallow = not actionable
• Focus on highest-leverage levels

Balance Leading and Lagging:

• Need both for complete picture
• Leading indicators for early action
• Lagging indicators for validation
• Don't optimize leading indicators that hurt lagging ones

Avoid Gaming:

• Consider unintended consequences
• What behaviors might teams game?
• Add guardrail metrics (quality, trust, safety)
• Balance growth with retention/satisfaction

Quick Reference

Resources:

• resources/template.md - Metrics tree structure with decomposition frameworks
• resources/methodology.md - Advanced techniques for complex metric systems
• resources/evaluators/rubric_metrics_tree.json - Quality criteria for metric trees

Output:

• File: metrics-tree.md in current directory
• Contains: North Star definition, input metrics (L2), action metrics (L3), leading indicators, prioritized experiments, metric relationships diagram

Success Criteria:

• North Star clearly defined with rationale
• 3-5 input metrics that fully decompose North Star
• Action metrics are specific, measurable behaviors
• Leading indicators identified with timing estimates
• Top 1-3 experiments prioritized with ICE scores
• Validated against rubric (score ≥ 3.5)

Quick Decision Framework:

• Simple product? → Use template.md with 2-3 levels
• Complex multi-sided? → Use methodology.md for separate trees per side
• Unsure about North Star? → Review common patterns above, test with "captures value + predicts revenue" criteria
• Too many metrics? → Limit to 3-5 per level, focus on highest impact

Common Mistakes:

1. Choosing wrong North Star: Pick vanity metric or one team can't influence
2. Too many levels: Analysis paralysis, lose actionability
3. Weak causal links: Metrics correlated but not causally related
4. Ignoring tradeoffs: Optimizing one metric hurts another
5. No experiments: Build tree but don't test hypotheses

Metrics Tree Methodology

When to use this methodology: You've used template.md and need advanced techniques for:

• Multi-sided marketplaces or platforms
• Complex metric interdependencies and feedback loops
• Counter-metrics and guardrail systems
• Network effects and viral growth
• Preventing metric gaming
• Seasonal adjustment and cohort aging effects
• Portfolio approach for different business stages

If your metrics tree is straightforward: Use template.md directly. This methodology is for complex metric systems.

---

Table of Contents

1. Multi-Sided Marketplace Metrics
2. Counter-Metrics & Guardrails
3. Network Effects & Viral Loops
4. Preventing Metric Gaming
5. Advanced Leading Indicators
6. Metric Interdependencies
7. Business Stage Metrics

---

1. Multi-Sided Marketplace Metrics

Challenge

Marketplaces have supply-side and demand-side that must be balanced. Optimizing one side can hurt the other.

Solution: Dual Tree Approach

Step 1: Identify constraint

• Supply-constrained: More demand than supply → Focus on supply-side metrics
• Demand-constrained: More supply than demand → Focus on demand-side metrics
• Balanced: Need both → Monitor ratio/balance metrics

Step 2: Create separate trees

Supply-Side Tree:

North Star: Active Suppliers (providing inventory)
├─ New supplier activation
├─ Retained suppliers (ongoing activity)
└─ Supplier quality/performance

Demand-Side Tree:

North Star: Successful Transactions
├─ New customer acquisition
├─ Repeat customer rate
└─ Customer satisfaction

Step 3: Define balance metrics

• Liquidity ratio: Supply utilization rate (% of inventory sold)
• Match rate: % of searches resulting in transaction
• Wait time: Time from demand signal to fulfillment

Example (Uber):

• Supply NS: Active drivers with >10 hours/week
• Demand NS: Completed rides
• Balance metric: Average wait time <5 minutes, driver utilization >60%

Multi-Sided Decomposition Template

Marketplace GMV = (Supply × Utilization) × (Demand × Conversion) × Average Transaction

Where:
- Supply: Available inventory/capacity
- Utilization: % of supply that gets used
- Demand: Potential buyers/searches
- Conversion: % of demand that transacts
- Average Transaction: $ per transaction

---

2. Counter-Metrics & Guardrails

Problem

Optimizing primary metrics can create negative externalities (quality drops, trust declines, user experience suffers).

Solution: Balanced Scorecard with Guardrails

Framework:

1. Primary metric (North Star): What you're optimizing
2. Counter-metrics: What could be harmed
3. Guardrail thresholds: Minimum acceptable levels

Example (Content Platform):

Primary: Content Views (maximize)

Counter-metrics with guardrails:
- Content quality score: Must stay ≥7/10 (current: 7.8)
- User satisfaction (NPS): Must stay ≥40 (current: 52)
- Creator retention: Must stay ≥70% (current: 75%)
- Time to harmful content takedown: Must be ≤2 hours (current: 1.5h)

Rule: If any guardrail is breached, pause optimization of primary metric

Common Counter-Metric Patterns

How to Set Guardrails

1. Historical baseline: Look at metric over past 6-12 months, set floor at 10th percentile
2. Competitive benchmark: Set floor at industry average
3. User feedback: Survey users on acceptable minimum
4. Regulatory: Compliance thresholds

---

3. Network Effects & Viral Loops

Measuring Network Effects

Network effect: Product value increases as more users join.

Metrics to track:

• Network density: Connections per user (higher = stronger network)
• Cross-side interactions: User A's action benefits User B
• Viral coefficient (K): New users generated per existing user
  • K > 1: Exponential growth (viral)
  • K < 1: Sub-viral (need paid acquisition)

Decomposition:

New Users = Existing Users × Invitation Rate × Invitation Acceptance Rate

Example:
100,000 users × 2 invites/user × 50% accept = 100,000 new users (K=1.0)

Viral Loop Metrics Tree

North Star: Viral Coefficient (K)

Decomposition:

K = (Invitations Sent / User) × (Acceptance Rate) × (Activation Rate)

Input metrics:
├─ Invitations per user
│  ├─ % users who send ≥1 invite
│  ├─ Average invites per sender
│  └─ Invitation prompts shown
├─ Invite acceptance rate
│  ├─ Invite message quality
│  ├─ Social proof (sender credibility)
│  └─ Landing page conversion
└─ New user activation rate
   ├─ Onboarding completion
   ├─ Value realization time
   └─ Early engagement actions

Network Density Metrics

Measure connectedness:

• Average connections per user
• % of users with ≥N connections
• Clustering coefficient (friends-of-friends)
• Active daily/weekly connections

Threshold effects:

• Users with 7+ friends have 10x retention (identify critical mass)
• Teams with 10+ members have 5x engagement (team size threshold)

---

4. Preventing Metric Gaming

Problem

Teams optimize for the letter of the metric, not the spirit, creating perverse incentives.

Gaming Detection Framework

Step 1: Anticipate gaming For each metric, ask: "How could someone game this?"

Example metric: Time on site

• Gaming: Auto-play videos, infinite scroll, fake engagement
• Intent: User finds value, willingly spends time

Step 2: Add quality signals Distinguish genuine value from gaming:

Time on site (primary)
+ Quality signals (guards against gaming):
  - Active engagement (clicks, scrolls, interactions) vs passive
  - Return visits (indicates genuine interest)
  - Completion rate (finished content vs bounced)
  - User satisfaction rating
  - Organic shares (not prompted)

Step 3: Test for gaming

• Spot check: Sample user sessions, review for patterns
• Anomaly detection: Flag outliers (10x normal behavior)
• User feedback: "Was this session valuable to you?"

Gaming Prevention Patterns

Pattern 1: Combination metrics Don't measure single metric; require multiple signals:

❌ Single: Pageviews
✓ Combined: Pageviews + Time on page >30s + Low bounce rate

Pattern 2: User-reported value Add subjective quality check:

Primary: Feature adoption rate
+ Counter: "Did this feature help you?" (must be >80% yes)

Pattern 3: Long-term outcome binding Tie short-term to long-term:

Primary: New user signups
+ Bound to: 30-day retention (signups only count if user retained)

Pattern 4: Peer comparison Normalize by cohort or segment:

Primary: Sales closed
+ Normalized: Sales closed / Sales qualified leads (prevents cherry-picking easy wins)

---

5. Advanced Leading Indicators

Technique 1: Propensity Scoring

Predict future behavior from early signals.

Method:

1. Collect historical data: New users + their 30-day outcomes
2. Identify features: Day 1 behaviors (actions, time spent, features used)
3. Build model: Logistic regression or decision tree predicting 30-day retention
4. Score new users: Probability of retention based on day 1 behavior
5. Threshold: Users with >70% propensity score are "likely retained"

Example (SaaS):

30-day retention model (R² = 0.78):
Retention = 0.1 + 0.35×(invited teammate) + 0.25×(completed 3 workflows) + 0.20×(time in app >20min)

Leading indicator: % of users with propensity score >0.7
Current: 45% → Target: 60% (predicts 15% retention increase)

Technique 2: Cohort Behavior Clustering

Find archetypes that predict outcomes.

Method:

1. Segment users by first-week behavior patterns
2. Measure long-term outcomes per segment
3. Identify high-value archetypes

Example:

Archetypes (first week):
- "Power user": 5+ days active, 20+ actions → 85% retain
- "Social": Invites 2+ people, comments 3+ times → 75% retain
- "Explorer": Views 10+ pages, low actions → 40% retain
- "Passive": <3 days active, <5 actions → 15% retain

Leading indicator: % of new users becoming "Power" or "Social" archetypes
Target: Move 30% → 45% into high-value archetypes

Technique 3: Inflection Point Analysis

Find tipping points where behavior changes sharply.

Method:

1. Plot outcome (retention) vs candidate metric (actions taken)
2. Find where curve steepens (inflection point)
3. Set that as leading indicator threshold

Example:

Retention by messages sent (first week):
- 0-2 messages: 20% retention (slow growth)
- 3-9 messages: 45% retention (moderate growth)
- 10+ messages: 80% retention (sharp jump)

Inflection point: 10 messages
Leading indicator: % of users hitting 10+ messages in first week

---

6. Metric Interdependencies

Problem

Metrics aren't independent; changing one affects others in complex ways.

Solution: Causal Diagram

Step 1: Map relationships Draw arrows showing how metrics affect each other:

[Acquisition] → [Active Users] → [Engagement] → [Retention]
     ↓                                              ↑
[Activation] ----------------------------------------

Step 2: Identify feedback loops

• Positive loop (reinforcing): A → B → A (exponential) Example: More users → more network value → more users
• Negative loop (balancing): A → B → ¬A (equilibrium) Example: More supply → lower prices → less supply

Step 3: Predict second-order effects If you increase metric X by 10%:

• Direct effect: Y increases 5%
• Indirect effect: Y affects Z, which feeds back to X
• Net effect: May be amplified or dampened

Example (Marketplace):

Increase driver supply +10%:
  → Wait time decreases -15%
  → Rider satisfaction increases +8%
  → Rider demand increases +5%
  → Driver earnings decrease -3% (more competition)
  → Driver churn increases +2%
  → Net driver supply increase: +10% -2% = +8%

Modeling Tradeoffs

Technique: Regression or experiments

Run A/B test increasing X
Measure all related metrics
Calculate elasticities:
  - If X increases 1%, Y changes by [elasticity]%
Build tradeoff matrix

Tradeoff Matrix Example:

Interpretation: Investing in retention has best ROI (15% revenue lift vs 6% from acquisition).

---

7. Business Stage Metrics

Problem

Optimal metrics change as business matures. Early-stage metrics differ from growth or maturity stages.

Stage-Specific North Stars

Pre-Product/Market Fit (PMF):

• Focus: Finding PMF, not scaling
• North Star: Retention (evidence of value)
• Key metrics:
  • Week 1 → Week 2 retention (>40% = promising)
  • NPS or "very disappointed" survey (>40% = good signal)
  • Organic usage frequency (weekly+ = habit-forming)

Post-PMF, Pre-Scale:

• Focus: Unit economics and growth
• North Star: New activated users per week (acquisition + activation)
• Key metrics:
  • LTV/CAC ratio (target >3:1)
  • Payback period (target <12 months)
  • Month-over-month growth rate (target >10%)

Growth Stage:

• Focus: Efficient scaling
• North Star: Revenue or gross profit
• Key metrics:
  • Net revenue retention (target >100%)
  • Magic number (ARR growth / S&M spend, target >0.75)
  • Burn multiple (cash burned / ARR added, target <1.5)

Maturity Stage:

• Focus: Profitability and market share
• North Star: Free cash flow or EBITDA
• Key metrics:
  • Operating margin (target >20%)
  • Market share / competitive position
  • Customer lifetime value

Transition Triggers

When to change North Star:

PMF → Growth: When retention >40%, NPS >40, organic growth observed
Growth → Maturity: When growth rate <20% for 2+ quarters, market share >30%

Migration approach:

1. Track both old and new North Star for 2 quarters
2. Align teams on new metric
3. Deprecate old metric
4. Update dashboards and incentives

---

Quick Decision Trees

"Should I use counter-metrics?"

Is primary metric easy to game or has quality risk?
├─ YES → Add counter-metrics with guardrails
└─ NO → Is metric clearly aligned with user value?
    ├─ YES → Primary metric sufficient, monitor only
    └─ NO → Redesign metric to better capture value

"Do I have network effects?"

Does value increase as more users join?
├─ YES → Track network density, K-factor, measure at different scales
└─ NO → Does one user's action benefit others?
    ├─ YES → Measure cross-user interactions, content creation/consumption
    └─ NO → Standard metrics tree (no network effects)

"Should I segment my metrics tree?"

Do different user segments have different behavior patterns?
├─ YES → Do segments have different value to business?
    ├─ YES → Create separate trees per segment, track segment mix
    └─ NO → Single tree, annotate with segment breakdowns
└─ NO → Are there supply/demand sides?
    ├─ YES → Dual trees (Section 1)
    └─ NO → Single unified tree

---

Summary: Advanced Technique Selector

Metrics Tree Template

How to Use This Template

Follow this structure to create a metrics tree for your product or business:

1. Start with North Star metric definition
2. Apply appropriate decomposition method
3. Map action metrics for each input
4. Identify leading indicators
5. Prioritize experiments using ICE framework
6. Output to metrics-tree.md

---

Part 1: North Star Metric

Define Your North Star

North Star Metric: [Name of metric]

Definition: [Precise definition including time window] Example: "Number of unique users who complete at least one transaction per week"

Rationale: [Why this metric?]

• ✓ Captures value delivered to customers: [how]
• ✓ Reflects business model: [revenue connection]
• ✓ Measurable and trackable: [data source]
• ✓ Actionable by teams: [who can influence]

Current Value: [Number] as of [Date]

Target: [Goal] by [Date]

North Star Selection Checklist

• Customer value: Does it measure value delivered to customers?
• Business correlation: Does it predict revenue/business success?
• Actionable: Can teams influence it through their work?
• Measurable: Do we have reliable data?
• Not vanity: Does it reflect actual usage/value, not just interest?
• Time-bounded: Does it have a clear time window (daily/weekly/monthly)?

---

Part 2: Metric Decomposition

Choose the decomposition method that best fits your North Star:

Method 1: Additive Decomposition

Use when: North Star is sum of independent segments

Formula:

North Star = Component A + Component B + Component C + ...

Template:

[North Star] =
  + [New users/customers]
  + [Retained users/customers]
  + [Resurrected users/customers]
  + [Other segment]

Example (SaaS WAU):

Weekly Active Users =
  + New activated users this week (30%)
  + Retained from previous week (60%)
  + Resurrected (inactive→active) (10%)

Method 2: Multiplicative Decomposition

Use when: North Star is product of rates/factors

Formula:

North Star = Factor A × Factor B × Factor C × ...

Template:

[North Star] =
  [Total addressable users/visits]
  × [Conversion rate at step 1]
  × [Conversion rate at step 2]
  × [Value per conversion]

Example (E-commerce Revenue):

Monthly Revenue =
  Monthly site visitors
  × Purchase conversion rate (3%)
  × Average order value ($75)

Method 3: Funnel Decomposition

Use when: North Star is end of sequential conversion process

Formula:

North Star = Top of funnel → Step 1 → Step 2 → ... → Final conversion

Template:

[North Star] =
  [Total entries]
  × [Step 1 conversion %]
  × [Step 2 conversion %]
  × [Final conversion %]

Example (Paid SaaS Customers):

New paid customers/month =
  Free signups
  × Activation rate (complete onboarding) (40%)
  × Trial start rate (25%)
  × Trial→Paid conversion rate (20%)

Math: 1000 signups × 0.4 × 0.25 × 0.2 = 20 paid customers

Method 4: Cohort Decomposition

Use when: Retention is key driver, need to separate acquisition from retention

Formula:

North Star = Σ (Cohort Size_t × Retention Rate_t,n) for all cohorts

Template:

[North Star today] =
  [Users from Month 0] × [Month 0 retention rate]
  + [Users from Month 1] × [Month 1 retention rate]
  + ...
  + [Users from Month N] × [Month N retention rate]

Example (Subscription Service MAU):

March Active Users =
  Jan signups (500) × Month 2 retention (50%) = 250
  + Feb signups (600) × Month 1 retention (70%) = 420
  + Mar signups (700) × Month 0 retention (100%) = 700
  = 1,370 MAU

---

Part 3: Input Metrics (L2)

For each component in your decomposition, define as input metric:

Input Metric Template

Input Metric 1: [Name]

• Definition: [Precise definition]
• Current value: [Number]
• Target: [Goal]
• Owner: [Team/person]
• Relationship to North Star: [How it affects NS, with estimated coefficient] Example: "Increasing activation rate by 10% → 5% increase in WAU"

Input Metric 2: [Name] [Repeat for 3-5 input metrics]

Validation Questions

• Are all input metrics mutually exclusive? (No double-counting)
• Do they collectively exhaust the North Star? (Nothing missing)
• Can each be owned by a single team?
• Is each measurable with existing/planned instrumentation?
• Are they all at same level of abstraction?

---

Part 4: Action Metrics (L3)

For each input metric, identify specific user behaviors that drive it:

Action Metrics Template

For Input Metric: [Name of L2 metric]

Action 1: [Specific user behavior]

• Measurement: [How to track it]
• Frequency: [How often it happens]
• Impact: [Estimated effect on input metric]
• Current rate: [% of users doing this]

Action 2: [Another behavior] [Repeat for 3-5 actions per input]

Example (For input metric "Retained Users"):

Action 1: User completes core workflow

• Measurement: Track "workflow_completed" event
• Frequency: 5x per week average for active users
• Impact: Users with 3+ completions have 80% retention vs 20% baseline
• Current rate: 45% of users complete workflow at least once

Action 2: User invites teammate

• Measurement: "invite_sent" event with "invite_accepted" event
• Frequency: 1.2 invites per user on average
• Impact: Users who invite have 90% retention vs 40% baseline
• Current rate: 20% of users send at least one invite

---

Part 5: Leading Indicators

Identify early signals that predict North Star movement:

Leading Indicator Template

Leading Indicator 1: [Metric name]

• Definition: [What it measures]
• Timing: [How far in advance it predicts] Example: "Predicts week 4 retention"
• Correlation: [Strength of relationship] Example: "r=0.75 with 30-day retention"
• Actionability: [How teams can move it]
• Current value: [Number]

Example:

Leading Indicator: Day 1 Activation Rate

• Definition: % of new users who complete 3 key actions on first day
• Timing: Predicts 7-day and 30-day retention (measured day 1, predicts weeks ahead)
• Correlation: r=0.82 with 30-day retention. Users with Day 1 activation have 70% retention vs 15% without
• Actionability: Improve onboarding flow, reduce time-to-value, send activation nudges
• Current value: 35%

How to Find Leading Indicators

Method 1: Cohort analysis

• Segment users by early behavior (first day, first week)
• Measure long-term outcomes (retention, LTV)
• Find behaviors that predict positive outcomes

Method 2: Correlation analysis

• List all early-funnel metrics
• Calculate correlation with North Star or key inputs
• Select metrics with r > 0.6 and actionable

Method 3: High-performer analysis

• Identify users in top 20% for North Star metric
• Look at their first week/month behavior
• Find patterns that distinguish them from average users

---

Part 6: Experiment Prioritization

Use ICE framework to prioritize which metrics to improve:

ICE Scoring Template

Impact (1-10): How much will improving this metric affect North Star?

• 10 = Direct, large effect (e.g., 10% improvement → 8% NS increase)
• 5 = Moderate effect (e.g., 10% improvement → 3% NS increase)
• 1 = Small effect (e.g., 10% improvement → 0.5% NS increase)

Confidence (1-10): How certain are we about the relationship?

• 10 = Proven causal relationship with data
• 5 = Correlated, plausible causation
• 1 = Hypothesis, no data yet

Ease (1-10): How easy is it to move this metric?

• 10 = Simple change, 1-2 weeks
• 5 = Moderate effort, 1-2 months
• 1 = Major project, 6+ months

ICE Score = (Impact + Confidence + Ease) / 3

Prioritization Table

Top 3 Experiments

Experiment 1: [Name - highest ICE score]

• Hypothesis: [What we believe will happen]
• Metric to move: [Target metric]
• Expected impact: [Quantified prediction]
• Timeline: [Duration]
• Success criteria: [How we'll know it worked]

Experiment 2: [Second highest] [Repeat structure]

Experiment 3: [Third highest] [Repeat structure]

---

Part 7: Metric Relationships Diagram

Create visual representation of your metrics tree:

ASCII Tree Format

North Star: [Metric Name] = [Current Value]
│
├─ Input Metric 1: [Name] = [Value]
│  ├─ Action 1.1: [Behavior] = [Rate]
│  ├─ Action 1.2: [Behavior] = [Rate]
│  └─ Action 1.3: [Behavior] = [Rate]
│
├─ Input Metric 2: [Name] = [Value]
│  ├─ Action 2.1: [Behavior] = [Rate]
│  ├─ Action 2.2: [Behavior] = [Rate]
│  └─ Action 2.3: [Behavior] = [Rate]
│
└─ Input Metric 3: [Name] = [Value]
   ├─ Action 3.1: [Behavior] = [Rate]
   ├─ Action 3.2: [Behavior] = [Rate]
   └─ Action 3.3: [Behavior] = [Rate]

Leading Indicators:
→ [Indicator 1]: Predicts [what] by [timing]
→ [Indicator 2]: Predicts [what] by [timing]

Example (Complete Tree)

North Star: Weekly Active Users = 10,000
│
├─ New Activated Users = 3,000/week (30%)
│  ├─ Complete onboarding: 40% of signups
│  ├─ Connect data source: 25% of signups
│  └─ Invite teammate: 20% of signups
│
├─ Retained Users = 6,000/week (60%)
│  ├─ Use core feature 3+ times: 45% of users
│  ├─ Create content: 30% of users
│  └─ Engage with team: 25% of users
│
└─ Resurrected Users = 1,000/week (10%)
   ├─ Receive reactivation email: 50% open rate
   ├─ See new feature announcement: 30% click rate
   └─ Get @mentioned by teammate: 40% return rate

Leading Indicators:
→ Day 1 activation rate (35%): Predicts 30-day retention
→ 3 key actions in first session (22%): Predicts weekly usage

---

Output Format

Create metrics-tree.md with this structure:

# Metrics Tree: [Product/Business Name]

**Date:** [YYYY-MM-DD]
**Owner:** [Team/Person]
**Review Frequency:** [Weekly/Monthly]

## North Star Metric

**Metric:** [Name]
**Current:** [Value] as of [Date]
**Target:** [Goal] by [Date]
**Rationale:** [Why this metric]

## Decomposition Method

[Additive/Multiplicative/Funnel/Cohort]

**Formula:**
[Mathematical relationship]

## Input Metrics (L2)

### 1. [Input Metric Name]
- **Current:** [Value]
- **Target:** [Goal]
- **Owner:** [Team]
- **Impact:** [Effect on NS]

#### Actions (L3):
1. [Action 1]: [Current rate]
2. [Action 2]: [Current rate]
3. [Action 3]: [Current rate]

[Repeat for all input metrics]

## Leading Indicators

1. **[Indicator 1]:** [Definition]
   - Timing: [When it predicts]
   - Correlation: [Strength]
   - Current: [Value]

2. **[Indicator 2]:** [Definition]
   [Repeat structure]

## Prioritized Experiments

### Experiment 1: [Name] (ICE: [Score])
- **Hypothesis:** [What we believe]
- **Metric:** [Target]
- **Expected Impact:** [Quantified]
- **Timeline:** [Duration]
- **Success Criteria:** [Threshold]

[Repeat for top 3 experiments]

## Metrics Tree Diagram

[Include ASCII or visual diagram]

## Notes

- [Assumptions made]
- [Data gaps or limitations]
- [Next review date]

---

Quick Examples by Business Model

SaaS Example (Slack-style)

North Star: Teams sending 100+ messages per week

Decomposition (Additive):

Active Teams = New Active Teams + Retained Active Teams + Resurrected Teams

Input Metrics:

• New active teams: Complete onboarding + hit 100 messages in week 1
• Retained active teams: Hit 100 messages this week and last week
• Resurrected teams: Hit 100 messages this week but not last 4 weeks

Leading Indicators:

• 10 members invited in first day (predicts team activation)
• 50 messages sent in first week (predicts long-term retention)

E-commerce Example

North Star: Monthly Revenue

Decomposition (Multiplicative):

Revenue = Visitors × Purchase Rate × Average Order Value

Input Metrics:

• Monthly unique visitors (owned by Marketing)
• Purchase conversion rate (owned by Product)
• Average order value (owned by Merchandising)

Leading Indicators:

• Add-to-cart rate (predicts purchase)
• Product page views per session (predicts purchase intent)

Marketplace Example (Airbnb-style)

North Star: Nights Booked

Decomposition (Multi-sided):

Nights Booked = (Active Listings × Availability Rate) × (Searches × Booking Rate)

Input Metrics:

• Active host supply: Listings with ≥1 available night
• Guest demand: Unique searches
• Match rate: Searches resulting in booking

Leading Indicators:

• Host completes first listing (predicts long-term hosting)
• Guest saves listings (predicts future booking)