Growth analytics

Growth & Analytics

Guide founders from defining their first metrics (pre-launch) through sophisticated retention analysis and experimentation (post-launch).

Workflow

1. Diagnose Current State

Ask: "Where are you in your analytics journey?"

State	Signals	Next Step
Pre-launch	No users yet, needs to define what to track	→ Step 2: Metrics Framework
Early traction	Has users, unclear what metrics matter	→ Step 2: Metrics Framework
Tracking basics	Has metrics, needs North Star focus	→ Step 2: North Star Selection
Ready to experiment	Solid metrics, wants to run tests	→ Step 3: A/B Testing
Retention concerns	Users churning, needs analysis	→ Step 4: Retention & Cohorts
Dashboard needed	Wants visibility for team/investors	→ Step 5: Dashboard Design

2. Metrics Framework

Build a metrics system that drives the right behavior. See references/metrics-frameworks.md for complete framework library.

AARRR Pirate Metrics

The universal startup funnel framework:

Stage	Question	Example Metrics
Acquisition	How do users find you?	Visitors, signups, CAC by channel
Activation	Do they have a great first experience?	Completed onboarding, "aha moment" reached
Retention	Do they come back?	DAU/MAU, D1/D7/D30 retention, churn
Revenue	Do they pay?	Conversion rate, ARPU, LTV
Referral	Do they tell others?	NPS, referral rate, viral coefficient

Stage-Appropriate Focus:

Stage	Primary Focus	Why
Pre-PMF	Activation + Retention	Nothing else matters if product doesn't stick
Post-PMF	Revenue + Acquisition	Time to scale what works
Growth	All five, plus efficiency	Optimize the full funnel

North Star Metric

One metric that best captures core value delivered to customers.

Selection Criteria:

1. Measures value — Correlates with customers getting value
2. Leading indicator — Predicts future revenue/growth
3. Actionable — Team can influence it
4. Simple — Easy to understand and communicate

North Star Examples by Business Model:

Model	North Star	Why
B2B SaaS	Weekly Active Users, Features Used	Value = engagement with product
Marketplace	Transactions completed	Both sides getting value
Subscription	Weekly active subscribers	Retention predicts LTV
E-commerce	Repeat purchase rate	Loyalty = sustainable revenue
Usage-based	Monthly usage volume	Usage = revenue
Social/Consumer	DAU/MAU ratio	Engagement intensity

Supporting Metrics:

Every North Star needs 3-5 supporting metrics that explain HOW to move it:

North Star: Weekly Active Teams (B2B SaaS)
├── Activation: Teams completing onboarding
├── Engagement: Features used per team
├── Expansion: Seats added per team
└── Retention: Team churn rate

One Metric That Matters (OMTM)

For early-stage focus, pick ONE metric for a defined period:

OMTM Selection:

1. What's the biggest constraint right now?
2. What metric would prove that constraint is solved?
3. Can you move it in 4-8 weeks?

Examples:

• Pre-launch: "Waitlist signups" (validate demand)
• Beta: "D7 retention" (validate stickiness)
• Post-launch: "Activation rate" (validate onboarding)
• Growth: "Payback period" (validate unit economics)

3. A/B Testing & Experimentation

Run experiments that generate reliable insights. See references/ab-testing.md for templates and calculators.

Experiment Design Framework

Hypothesis Structure:

If we [change], then [metric] will [improve/decrease] by [amount]
because [reason based on user insight].

Example:

If we reduce signup form from 5 fields to 3 fields, then signup completion rate will increase by 15% because user research shows form length is the #1 drop-off reason.

Before Running Any Test

Pre-flight Checklist:

Check	Question	Action
Sample size	Do we have enough traffic?	Calculate minimum sample (see below)
Duration	How long to reach significance?	Usually 1-4 weeks minimum
Metric clarity	What exactly are we measuring?	Define primary + guardrail metrics
Segment impact	Should we segment results?	Pre-define segments (new vs returning, mobile vs desktop)

Sample Size Estimation:

For 80% power and 95% confidence:

• 10% baseline, detect 10% relative lift → ~15,000 per variant
• 10% baseline, detect 20% relative lift → ~4,000 per variant
• 2% baseline, detect 20% relative lift → ~20,000 per variant

Rule of thumb: Multiply expected traffic by test duration. If you can't reach minimum sample in 4 weeks, the test isn't worth running—make a bigger change.

Running the Test

Test Execution Rules:

1. Run for full weeks (capture day-of-week effects)
2. Don't peek early—commit to duration
3. Track guardrail metrics (what shouldn't break)
4. Document everything before launch

Guardrail Metrics Examples:

• Revenue per user (main metric might improve but hurt revenue)
• Page load time (change might slow performance)
• Support tickets (change might confuse users)

Interpreting Results

Result	Interpretation	Action
Significant win	p < 0.05, metric improved	Ship it, document learnings
Significant loss	p < 0.05, metric declined	Don't ship, learn why
Inconclusive	p > 0.05	Not enough data OR no real effect
Flat	Large sample, no movement	Effect likely too small to matter

Common Pitfalls:

• Stopping early when results look good (inflates false positives)
• Testing too many variants (dilutes sample)
• Ignoring segments (average hides important differences)
• No hypothesis (test without learning)

4. Retention & Cohort Analysis

Understand if users stick around. See references/retention-cohorts.md for SQL templates and benchmarks.

Retention Fundamentals

Types of Retention:

Type	Definition	Use When
N-day retention	% of users active on exactly day N	Daily-use products (social, games)
Bounded retention	% active within day range (e.g., week 1)	Weekly-use products (SaaS)
Unbounded retention	% active on day N or any day after	Long purchase cycles (e-commerce)

Critical Retention Points:

Timeframe	What It Measures	Healthy Benchmark
D1	First impression	>25% (consumer), >40% (B2B)
D7	Habit forming	>15% (consumer), >30% (B2B)
D30	Stickiness	>10% (consumer), >25% (B2B)
D90	Long-term value	Product-dependent

Cohort Analysis

Group users by signup date (or other dimension) to track behavior over time.

Cohort Table Structure:

Cohort	Week 0	Week 1	Week 2	Week 3	Week 4
Jan 1-7	100%	40%	30%	25%	22%
Jan 8-14	100%	45%	35%	28%	25%
Jan 15-21	100%	48%	38%	32%	—

Reading Cohort Tables:

• Rows = Compare cohorts (are newer users retaining better?)
• Columns = Retention decay (where's the biggest drop-off?)
• Diagonals = Same calendar week (external events)

Cohort Dimensions Beyond Time:

• Acquisition channel (organic vs. paid)
• Plan type (free vs. paid)
• First action taken (feature X vs. feature Y)
• Geography

Retention Curves

Healthy Curve Shape:

100% ─┐
      │╲
      │ ╲
      │  ╲____________________  ← Flattens = retention
      │
  0% ─┴─────────────────────────
      D1   D7   D30   D60   D90

Danger Signs:

• Curve never flattens (continuous bleed)
• Steep drop after D1 (activation problem)
• Drop at specific point (feature/billing issue)

Churn Analysis

Churn Rate Calculation:

Monthly Churn = Customers Lost This Month / Customers at Start of Month

Churn Benchmarks (SaaS):

Segment	Good	Great
SMB	<5% monthly	<3% monthly
Mid-market	<2% monthly	<1% monthly
Enterprise	<1% monthly	<0.5% monthly

Churn Diagnosis Questions:

1. When do they churn? (Tenure analysis)
2. Who churns? (Segment analysis)
3. Why do they churn? (Exit surveys, support tickets)
4. What predicts churn? (Behavioral signals)

5. Dashboard Design

Create visibility that drives action. See references/dashboard-design.md for templates and tool recommendations.

Dashboard Hierarchy

Level 1: Executive Dashboard (weekly, whole company)

• 3-5 top-level KPIs
• Trend vs. target
• One screen, no scrolling

Level 2: Functional Dashboards (daily, by team)

• Sales: Pipeline, conversion, activity
• Product: Engagement, retention, feature adoption
• Marketing: Acquisition, CAC, channel performance
• Support: Tickets, response time, CSAT

Level 3: Operational Dashboards (real-time, by function)

• Engineering: Uptime, latency, errors
• Sales: Daily activity, quota attainment

KPI Selection

For Each Metric, Answer:

1. What decision does this inform?
2. Who needs to see it and how often?
3. What's the target and why?
4. What action triggers if it's off-track?

Metric Types to Include:

Type	Purpose	Example
Leading	Predict future outcomes	Pipeline, activation rate
Lagging	Confirm results	Revenue, churn
Input	Activities you control	Calls made, features shipped
Output	Outcomes you want	Deals closed, retention

Visualization Principles

Choosing Chart Types:

Data Type	Best Chart
Trend over time	Line chart
Comparison across categories	Bar chart
Part-to-whole	Pie (if <5 segments), stacked bar
Distribution	Histogram
Correlation	Scatter plot
Funnel stages	Funnel chart

Dashboard Anti-Patterns:

• ❌ Too many metrics (more than 8-10 per view)
• ❌ No context (numbers without targets/trends)
• ❌ Vanity metrics (impressive but not actionable)
• ❌ Stale data (updated monthly when weekly needed)
• ❌ No owner (who acts on this?)

Tool Selection

Tool Recommendations by Stage:

Stage	Recommended Approach
Pre-launch	Spreadsheet (Google Sheets)
MVP/Beta	Simple analytics (Mixpanel free, Amplitude free, PostHog)
Post-PMF	Full stack (Mixpanel/Amplitude + data warehouse + BI tool)
Scaling	Custom (Segment → warehouse → Looker/Metabase)

Tool Comparison:

Tool	Best For	Limitation
Google Analytics	Web traffic, acquisition	Weak on product analytics
Mixpanel	Product analytics, funnels	Can get expensive at scale
Amplitude	Product analytics, cohorts	Learning curve
PostHog	Open source, self-hosted option	Younger product
Heap	Auto-capture everything	Data can be messy
Metabase	SQL-based, self-hosted BI	Requires data warehouse
Looker	Enterprise BI	Complex, expensive

6. Anti-Patterns

Metrics Mistakes:

• Tracking everything, focusing on nothing
• Vanity metrics (total signups vs. active users)
• Lagging-only metrics (revenue without leading indicators)
• No targets (data without context)

Experimentation Mistakes:

• Testing small changes on low-traffic pages
• Multiple changes in one test (can't isolate effect)
• Stopping tests early based on early results
• No hypothesis (random changes)

Retention Mistakes:

• Only looking at aggregate retention (hiding segment issues)
• Ignoring activation (retention starts at first experience)
• Not defining "active" clearly

Dashboard Mistakes:

• Dashboard nobody checks
• Real-time when weekly is sufficient
• No owners assigned to metrics

Deliverables

1. Metrics Framework Document

Create as markdown:

• North Star metric with rationale
• AARRR funnel with specific metrics
• Supporting metrics hierarchy
• Targets and owners

2. Metrics Tracker Spreadsheet

Create using xlsx skill:

• AARRR funnel metrics with weekly/monthly tracking
• Formulas for calculated metrics (conversion rates, growth rates)
• Target vs. actual comparison
• Charts for trends

3. A/B Test Plan

Create as markdown:

• Hypothesis statement
• Variants description
• Primary and guardrail metrics
• Sample size and duration calculation
• Success criteria

4. Cohort Analysis Spreadsheet

Create using xlsx skill:

• Cohort table (rows = cohorts, columns = time periods)
• Retention percentages with conditional formatting
• Retention curve visualization
• Cohort comparison charts

5. Dashboard Specification

Create as markdown:

• KPI hierarchy (executive → functional → operational)
• Metric definitions with formulas
• Visualization recommendations
• Data sources and refresh frequency
• Tool recommendation with rationale

6. SQL Query Templates

Create as markdown:

• Cohort retention query
• Funnel conversion query
• Active user calculation
• Churn identification query

Reference Files

• references/metrics-frameworks.md — AARRR deep dive, North Star selection guide, metrics by business model, anti-patterns
• references/ab-testing.md — Experiment templates, sample size calculator, significance interpretation, SQL queries
• references/retention-cohorts.md — Cohort methods, retention curves, SQL templates, benchmarks by model
• references/dashboard-design.md — Dashboard templates, visualization guide, tool comparison

Integration with Other Skills

• Use business-model skill for unit economics metrics (LTV, CAC, payback)
• Use product skill for feature prioritization based on analytics
• Use go-to-market skill for channel-specific acquisition metrics
• Use operations skill for OKRs aligned with metrics framework
• Use fundraising skill for investor-ready metrics presentation
• Use xlsx skill for metrics trackers and cohort spreadsheets
• Use docx skill for analytics documentation

---

Adapted from Linas Beliūnas's The One-Person Unicorn founder skill set.