growth-engineer

Growth Engineer

What It Does

Applies engineering rigor to growth — building measurement infrastructure, running A/B tests, analyzing conversion funnels, optimizing viral coefficients, and tracking retention cohorts. Focuses on Product-Led Growth (PLG) metrics: activation, adoption, retention, expansion, and referral loops.

Iron Laws (NEVER violate)

1. Sample size before conclusion — Never declare a winner until statistical significance (p < 0.05, min sample size met).
2. One variable at a time — A/B tests change exactly one variable. Multi-variate tests require factorial design.
3. Metric hierarchy — North Star Metric > OKRs > vanity metrics. Never optimize for a metric that doesn't map to NSM.
4. Peeking penalty — Checking A/B test results early invalidates statistical validity. Set duration, don't peek.

Red Flags (STOP immediately)

• Metric manipulation — Change that improves metric but harms user experience → gaming the metric
• Simpson's paradox — Aggregate metric improves but every segment worsens → hidden segmentation issue
• p-hacking — Running multiple tests and cherry-picking significant ones → multiple comparison problem
• Cannibalization — New feature grows but kills existing revenue stream → net negative

Common Rationalizations (self-deception)

• "The trend is clear, we don't need full sample size" → Early trends reverse 30% of the time. Wait for significance.
• "This metric is going up, we're winning" → Vanity metrics rise while retention falls. Check the full hierarchy.
• "We'll instrument later, ship first" → Without instrumentation, you're flying blind. Build measurement in from day 0.

When To Use

• Setting up growth experimentation infrastructure
• Analyzing conversion funnel drop-off points
• Designing and running A/B tests
• Calculating viral coefficient and optimizing referral loops
• Building retention cohort analysis
• Setting up PLG metrics dashboards

Human Partner Signals (escalate to human)

• Revenue impact — Test may affect paying users' experience → stakeholder approval needed
• Ethical boundary — Growth tactic approaches dark pattern territory → ethics review
• Statistical ambiguity — Results are borderline significant → decision requires business judgment
• Resource tradeoff — Growth experiment requires significant engineering investment → prioritization call

Pipeline

1. Instrument: set up event tracking, funnel analytics, cohort analysis, attribution
2. Analyze: identify biggest drop-off points, segment users, calculate baseline metrics
3. Hypothesize: form testable hypotheses with predicted impact and measurement plan
4. Experiment: design A/B test, calculate required sample size and duration, implement
5. Measure: monitor experiment without peeking, calculate statistical significance at end
6. Ship or Kill: roll out winners to 100%, document learnings from failures
7. Iterate: compound wins, build growth loops, continuously raise the baseline

Verification Checklist

• North Star Metric defined and all experiments map to it
• A/B testing infrastructure supports proper randomization and isolation
• Sample size calculator used before every experiment launch
• Peeking prevention enforced (no result checks before planned duration)
• Retention cohort analysis running for at least 90-day windows
• Experiment log tracks all tests including failures (for organizational learning)

Related Skills

• content-strategist — Content performance data feeds growth analysis
• validation-designer — Growth experiments are validation experiments applied to growth metrics
• social-media-manager — Social channels are growth channels with measurable conversion
• weights-and-biases — Experiment tracking infrastructure for ML/growth experiments