analytics-metrics-kpi（分析指标 kpi）

Analytics & Metrics Skill

Become data-driven. Define meaningful metrics, build dashboards, run experiments, and make decisions based on data, not intuition.

Metrics Framework (Acquisition → Revenue)

North Star Metric

Definition: One metric that best captures the value your product delivers.

Characteristics:

• Directly tied to business success
• Driven by product improvements
• Leading indicator of revenue
• Understandable to whole company

Examples:

• Slack: Daily Active Users (DAU)
• Airbnb: Booked Nights
• YouTube: Watch Time
• Uber: Rides Completed
• Stripe: Payment Volume Processed

Funnel Metrics (Acquisition)

Total Visitors: 100,000/month
↓ 20% conversion
Free Signups: 20,000
↓ 10% free-to-paid
Paid Customers: 2,000

CAC: $50 (marketing + sales spend / customers acquired)
LTCAC: $100 (all customer acquisition costs)

Metrics to Track:

• Traffic - Total visitors to website/app
• Signup Rate - % who sign up (target: 10-15%)
• Free-to-Paid Conversion - % free users who pay (target: 2-5%)
• CAC - Cost per acquired customer
• CAC Payback - Months to recover CAC from revenue (target: < 12 months)

Activation Metrics

Goal: New users become active users

Free Signups: 2,000
↓ 30% onboard successfully
Activated: 600
↓ 60% remain active Day 7
Day 7 Active: 360

Metrics to Track:

• Onboarding Completion Rate - % who complete setup (target: 50-80%)
• Time to First Value - Hours to first successful use
• Feature Adoption - % who try key features
• Day 1/7/30 Retention - % active those days (target: 40/25/15)

Engagement Metrics

Goal: Users regularly use product

Daily/Monthly Metrics:

• DAU/MAU - Daily/Monthly Active Users
• DAU/MAU Ratio - Stickiness (target: 20-30%)
• Feature Usage - % using key features
• Session Length - Minutes per session
• Session Frequency - Times per week

Cohort Analysis Example:

Jan Cohort (1,000 signups):
- Day 1: 600 active (60%)
- Day 7: 360 active (36%)
- Day 30: 180 active (18%)
- Month 3: 90 active (9%)

Feb Cohort (1,500 signups):
- Day 1: 1050 active (70%) ← Improving!
- Day 7: 630 active (42%)
- Day 30: 300 active (20%)

Retention Metrics

Goal: Users stay and continue paying

Month 1: 1,000 customers
Month 2: 900 active (90% retained)
Month 3: 810 active (90% of month 2)
Month 12: 314 active (31% annual retention)

Churn Rate: % lost each period

• Monthly churn: (Customers Lost / Month Start) × 100
• Annual churn: 1 - (Ending / Starting)
• Target for SaaS: < 5% monthly churn

NPS (Net Promoter Score)

• Question: "How likely to recommend (0-10)?"
• Score = % Promoters (9-10) - % Detractors (0-6)
• Range: -100 to +100
• Target: 50+ (world-class)

Revenue Metrics

Monthly Recurring Revenue (MRR)

MRR = (Total paid customers) × (average subscription price)
Growth MRR = New MRR + Expansion MRR - Churn MRR

Annual Run Rate (ARR)

ARR = MRR × 12

Average Revenue Per User (ARPU)

ARPU = MRR / Total Users

Customer Lifetime Value (LTV)

LTV = (ARPU × Gross Margin %) / Monthly Churn %

Example:
ARPU: $100
Gross Margin: 80%
Monthly Churn: 5%
LTV = ($100 × 80%) / 5% = $1,600

If CAC = $400: LTV/CAC = 4x ✓ (target: 3x+)

Dashboard Architecture

Executive Dashboard (C-Level)

Weekly Updates:

• MRR / ARR (vs target, vs month ago)
• New customers (weekly, monthly)
• Churn rate (%)
• NPS score
• Engagement (DAU, MAU)
• Key initiatives status

Frequency: Weekly

Product Dashboard (Product Team)

Daily/Weekly:

• Funnel metrics (signup → paid)
• Feature adoption
• Engagement metrics
• User feedback score
• A/B test results
• Support ticket volume

Frequency: Daily updates

Financial Dashboard (Finance/Operations)

Monthly:

• MRR / ARR
• Customer acquisition cost
• Customer lifetime value
• Gross margin
• CAC payback period
• Revenue by segment
• Churn by cohort

Frequency: Monthly

Health Dashboard (Operations)

Realtime:

• System uptime (%)
• Error rate (%)
• Response time (p95)
• Database performance
• Support ticket response time
• Support backlog

Frequency: Realtime/hourly

A/B Testing (Experimentation)

Test Planning

Hypothesis: "If we change X, then Y will improve, because Z"

Example: "If we move signup button above the fold, then conversion will improve 15%, because users won't scroll."

Test Structure

Experiment Design:

• Control: Keep current version
• Treatment: New version
• Sample size: Enough users to be statistical
• Duration: 2-4 weeks minimum
• Metric: Clear success metric

Statistical Significance

Confidence Level: 95% (industry standard)

• Means 5% chance of false positive
• Need enough samples (typically 1000-10K per variant)
• Use calculator for exact sample size

P-Value: Probability result is random chance

• P < 0.05: Statistically significant
• P > 0.05: Not significant, inconclusive

Example A/B Test

Hypothesis: Moving signup button above fold increases conversion 15%

Setup:

• Control: Current design
• Treatment: Button moved above fold
• Success metric: Conversion rate (signup / visit)
• Sample size: 10,000 users per variant
• Duration: 2 weeks
• Confidence: 95%

Results:

• Control: 2.0% conversion (200 signups from 10K visitors)
• Treatment: 2.8% conversion (280 signups from 10K visitors)
• Improvement: 40% increase (0.8% / 2% = 40%)
• P-value: 0.02 (statistically significant!)
• Decision: SHIP IT - Roll out to 100%

Test Ideas by Priority

High Priority (Start Here):

• Signup flow optimization (biggest funnel)
• Onboarding experience
• Pricing page clarity
• Feature discoverability

Medium Priority:

• UI copy optimization
• CTA button colors
• Email subject lines
• Notification triggers

Low Priority:

• Micro-copy tweaks
• Animation effects
• Color scheme changes

Metric Pitfalls to Avoid

Vanity Metrics

❌ "We have 1M page views!" ✓ "We have 50K daily active users, growing 10% monthly"

Actionable vs Non-Actionable

❌ "User satisfaction increased" (what changed?) ✓ "Onboarding completion rate 65% → 78% (↑20%)" (clear action)

Correlation vs Causation

❌ "Ice cream sales correlate with drownings" ✓ Understand actual causation, not just correlation

Look-Alike Metrics

❌ Track MRR but not Customer LTV (can grow MRR by spending more on acquisition) ✓ Track both acquisition efficiency AND retention

Metrics Review Cadence

Daily:

• System uptime
• Error rates
• Support response time

Weekly:

• Funnel metrics
• Feature adoption
• Key engagement metrics
• Test results

Monthly:

• Revenue metrics
• Cohort analysis
• Churn breakdown
• LTV/CAC trends

Quarterly:

• Strategic metric review
• Long-term trend analysis
• Metric changes needed

Troubleshooting

Yaygın Hatalar & Çözümler

Debug Checklist

[ ] North Star metric defined mi?
[ ] Metrics business goals'a aligned mi?
[ ] Data collection accurate mi?
[ ] Dashboard refreshed mi?
[ ] A/B test sample sufficient mi?
[ ] Statistical significance achieved mi?

Recovery Procedures

1. Data Quality Issues → Flag affected metrics, exclude
2. Inconclusive A/B → Extend test duration
3. Misleading Metrics → Add context/segmentation

Master data-driven decision making and grow faster!