Growth & Analytics

Core Principle

If you can't measure it, you can't improve it. But measuring everything improves nothing. Pick the metrics that drive decisions, ignore the rest.

AARRR Pirate Metrics Framework

The Funnel

ACQUISITION  → How do users find you?
ACTIVATION   → Do they have a great first experience?
RETENTION    → Do they come back?
REVENUE      → Do they pay you?
REFERRAL     → Do they tell others?

Metrics by Stage

Where to Focus First

Fix the funnel from the BOTTOM UP:
  1. Retention first — no point acquiring users who leave
  2. Activation second — make the first experience great
  3. Acquisition third — now pour users into a funnel that works
  4. Revenue fourth — monetize retained users
  5. Referral last — happy retained users refer naturally

North Star Metric Selection

What Makes a Good North Star

A North Star metric must:
  1. Reflect core product value (not vanity)
  2. Be a leading indicator of revenue
  3. Be actionable by the team
  4. Be understandable by everyone

Examples by Company Type

KPI vs. OKR — The Distinction

KPIs = What you WATCH (ongoing health metrics)
  "MRR is $50K" — it tells you the state of things

OKRs = What you CHANGE (time-bound improvement targets)
  "Increase MRR from $50K to $80K by Q2" — it drives action

KPIs without OKRs = dashboard watching
OKRs without KPIs = flying blind

A/B Testing

Sample Size Calculation

Required sample per variant:
  n = (Z² × p × (1-p)) / E²

Where:
  Z = 1.96 (95% confidence)
  p = baseline conversion rate
  E = minimum detectable effect

Example:
  Baseline conversion: 5% (p = 0.05)
  Want to detect 20% relative lift (5% → 6%, so E = 0.01)
  n = (1.96² × 0.05 × 0.95) / 0.01² = 1,825 per variant
  Total needed: 3,650 visitors

Testing Rules

1. One variable at a time — otherwise you can't attribute results
2. Run to full sample size — don't peek and stop early
3. Statistical significance ≠ practical significance — a 0.1% lift
   at p<0.05 isn't worth shipping
4. Test the big levers first — headlines, CTAs, pricing, not button colors
5. Document every test — hypothesis, result, learning

Testing Prioritization

Retention & Cohort Analysis

The Core Insight

A healthy retention curve FLATTENS.
An unhealthy retention curve approaches zero.

Healthy:  Week 1: 100% → Week 4: 45% → Week 8: 38% → Week 12: 35%
                                                        ↑ Flattening

Unhealthy: Week 1: 100% → Week 4: 30% → Week 8: 15% → Week 12: 5%
                                                        ↑ Still declining

Cohort Analysis SQL Template

-- Weekly retention cohorts
WITH cohort AS (
  SELECT
    user_id,
    DATE_TRUNC('week', MIN(created_at)) AS cohort_week
  FROM events
  WHERE event_name = 'signup'
  GROUP BY user_id
),
activity AS (
  SELECT
    user_id,
    DATE_TRUNC('week', event_time) AS active_week
  FROM events
  WHERE event_name = 'core_action'
  GROUP BY user_id, DATE_TRUNC('week', event_time)
)
SELECT
  c.cohort_week,
  DATEDIFF('week', c.cohort_week, a.active_week) AS week_number,
  COUNT(DISTINCT a.user_id) AS active_users,
  COUNT(DISTINCT a.user_id)::FLOAT
    / COUNT(DISTINCT c.user_id) AS retention_rate
FROM cohort c
LEFT JOIN activity a ON c.user_id = a.user_id
  AND a.active_week >= c.cohort_week
GROUP BY c.cohort_week, week_number
ORDER BY c.cohort_week, week_number;

Reading the Cohort Table

          Week 0  Week 1  Week 2  Week 4  Week 8
Jan cohort  100%    52%     41%     35%     33%   ← Healthy (flattening)
Feb cohort  100%    48%     30%     18%     10%   ← Problem (still declining)
Mar cohort  100%    55%     45%     40%     38%   ← Best (improved onboarding)

Compare cohorts to see if product changes improve retention.

Anti-Patterns

Power Move

"Set up a growth analytics framework for [product]. Define the North Star metric, map AARRR metrics with targets, and write a cohort analysis SQL query for our retention data. Then identify the biggest leak in the funnel and design an A/B test to fix it."

The agent becomes your growth analyst — turning data into decisions and experiments into compounding improvements.