Metrics Framework: Leading vs Lagging Indicators

When to use: Setting up metrics, designing dashboards, or when you need early signal before outcomes materialize

Framework source: Aakash Gupta's "Become an A/B Testing Expert" and metrics guidance

Quick Start

1. Tell me: "Help me build a metrics framework for [product/feature]"
2. I will check thoughts/shared/pm/frameworks/ for your North Star and thoughts/shared/pm/metrics/ for baseline data
3. I will ask about your North Star (lagging metric), product stage, and decision speed needs
4. We build a metric hierarchy: Lagging (quarterly) -> Leading (weekly) -> Input (daily)
5. For each key metric, we define alert thresholds (green/yellow/red) and a dashboard layout
6. Output goes to thoughts/shared/pm/metrics/metrics-framework-[date].md

Key principle: Leading metrics should PREDICT lagging metrics. If they do not, you are tracking the wrong thing. We always validate the correlation before committing to a leading indicator.

Context Routing Logic (Internal - for Claude)

Automatic Context Checks: When this skill is invoked, immediately check:

Context Priority:

1. Business strategy and North Star FIRST
2. Current product metrics and baselines SECOND
3. Historical trend data THIRD
4. Feature-specific metrics FOURTH

Cross-Skill Links:

• If defining North Star → Link to /define-north-star
• If setting up feature success → Link to /feature-metrics for STEDII framework
• If tracking product health → Link to /retention-analysis and /activation-analysis
• If analyzing test results → Link to /feature-results

Step 0: Understanding Your Metric System

Before designing your metrics framework, let me understand what you're optimizing for...

Checking:

• thoughts/shared/pm/context/business-info-template.md for business model and goals
• thoughts/shared/pm/frameworks/*.md for strategic priorities
• thoughts/shared/pm/metrics/ for existing metric baselines
• thoughts/shared/pm/prds/ for current feature success metrics

Based on what I find, I'll show you:

Your Current Metric Health

North Star / Lagging Metric:

• [Current North Star metric from strategy]
• [Baseline value: X]
• [Target: Y by when]

Existing Leading Indicators:

• [What you're already tracking]
• [Historical effectiveness: does this predict the lagging metric?]

Metric Gaps:

• [Metrics you're missing that would help you move faster]
• [Decision-making gaps: where you're flying blind]

PM-Specific Diagnosis Questions

1. Decision Speed: How long do you need to make go/no-go decisions? (Days vs weeks vs months)
2. Product Stage: Are you focused on growth, retention, monetization, or operational efficiency?
3. Team Alignment: Do different teams use different success metrics? (Problem!)
4. Data Quality: Do you have reliable tracking for the metrics you care about?
5. Stakeholder Expectations: Does your executive team care about your current metrics?

Leading vs Lagging Metrics

Lagging Metrics (Outcome Indicators)

Definition: Metrics that measure final outcomes - what already happened

Characteristics:

• Slow to respond
• High certainty (accurate)
• Ultimate measure of success
• Hard to influence directly

Examples:

• Revenue
• Customer lifetime value (LTV)
• Annual retention rate
• Net Promoter Score (NPS)
• Market share
• Profit margins

When to use:

• Board reporting
• Annual/quarterly goals
• Validating long-term bets
• Measuring ultimate success

Leading Metrics (Predictive Indicators)

Definition: Metrics that predict future outcomes - early signals

Characteristics:

• Fast to respond
• Lower certainty (probabilistic)
• Actionable in real-time
• Directly influenceable

Examples:

• Day 7 activation rate (predicts retention)
• Weekly active users (predicts revenue)
• Feature adoption rate (predicts stickiness)
• Time to first value (predicts churn)
• Support ticket volume (predicts satisfaction)

When to use:

• Day-to-day decisions
• Experiment evaluation
• Early problem detection
• Sprint planning

The Connection

Leading metrics should PREDICT lagging metrics.

If they don't, you're tracking the wrong leading metrics.

Example 1: E-commerce

Lagging: Monthly revenue Leading:

• Add-to-cart rate (faster signal)
• Checkout start rate
• Payment success rate

Logic: If add-to-cart drops today, revenue drops next week.

Example 2: SaaS Product

Lagging: 90-day retention Leading:

• Day 7 activation rate
• Weekly active usage
• Feature adoption in first 14 days

Logic: Users who activate in 7 days have 10x better 90-day retention.

Example 3: Marketplace

Lagging: GMV (Gross Merchandise Value) Leading:

• Search-to-view rate
• View-to-cart rate
• Seller response time
• Inventory depth

Logic: More searches + better inventory = more transactions = higher GMV

How to Identify Your Leading Metrics

Step 1: Start with Your North Star (Lagging Metric)

What's the ultimate outcome you care about?

Examples:

• Revenue
• User retention
• Engagement
• Customer satisfaction

Step 2: Work Backwards to Find Predictors

Use this prompt pattern:

Use /metrics-framework and reference thoughts/shared/pm/context/business-info-template.md

Our North Star / main outcome metric is: [your lagging metric]

Help me identify 3-5 leading indicators that:
1. Happen BEFORE this outcome
2. Predict this outcome with data
3. Are actionable (we can influence them)
4. Give us signal within 1-2 weeks

Our product: [describe product]
User journey: [describe key actions]
Current metrics: [what you track today]

Step 3: Validate the Correlation

Run analysis to prove the connection:

1. Cohort analysis:

   • Users with high leading metric → retention/revenue outcome
   • Users with low leading metric → churn outcome

2. Time-series correlation:

   • Leading metric moves up Week 1
   • Lagging metric moves up Week 4-6
   • Correlation coefficient > 0.7

If no correlation exists, it's not a leading indicator.

Minimum Data Quality Bar

Before trusting any correlation between leading and lagging metrics:

Minimum thresholds:

• Sample size: At least 1,000 users (or 100 accounts for B2B) in each comparison group
• Time period: At least 6 weeks of data (2 full product cycles minimum)
• Event frequency: Each user should have 5+ events of the measured behavior (avoid drawing conclusions from single actions)
• Segment stability: The user segments being compared should be stable (not changing composition) during the measurement period

Data quality red flags:

• Correlation based on fewer than 500 users -- mark as "Directional only, not statistically reliable"
• Correlation from a single week of data -- mark as "Preliminar