Growth Engineering & Funnel Optimization
Overview: Growth engineering treats the funnel as a system — stages, conversion rates, guardrails, and feedback loops — improved through prioritized experiments rather than one-off campaigns. The goal is repeatable…
AARRR (“pirate metrics”) framework
| Stage | Definition | Key metrics (examples) | Example KPIs | Typical ownership |
|---|---|---|---|---|
| Acquisition | Users arrive via channels | CAC, installs, signups, traffic by source | Blended CAC, paid vs organic mix, channel LTV | Marketing, Growth |
| Activation | Users reach first meaningful value | Time-to-value, activation rate, onboarding completion | % completing “aha” event in session 1 | Product, Growth |
| Retention | Users return or stay subscribed | D1/D7/D30, churn, cohort curves | Net revenue retention (NRR), WAU/MAU | Product, CS, Growth |
| Revenue | Monetization and expansion | ARPU, expansion, trial-to-paid | LTV, paywall conversion, upgrade rate | Product, Finance, Growth |
| Referral | Users bring others | K-factor (use carefully), invite rate, NPS → referral | Viral coefficient by cohort, referral % of signups | Growth, Product |
AARRR funnel with conversion rates
Growth process cycle
- ICE / RICE: Score ideas on Impact, Confidence, Ease (ICE) or add Reach for RICE — use consistently so the backlog is comparable week to week.
- Learning: Document hypothesis, design, result, decision — even null results reduce duplicate work.
Funnel instrumentation and guardrails
| Layer | Examples |
|---|---|
| Stage definitions | Mutually exclusive rules (e.g. “activated” = completed event X within Y hours of signup) |
| Event quality | Schema versioning, deduplication, identity stitching for logged-in vs anonymous |
| Guardrail metrics | Refund rate, chargebacks, support tickets per thousand users, spam signups — must not regress while optimizing conversion |
| Qualitative triangulation | Session replay (privacy-safe), interviews, and support themes to explain why a metric moved |
Without guardrails, teams often “win” experiments that damage trust, compliance, or operational load — then pay the cost in the next quarter.
Acquisition optimization
| Lever | Notes |
|---|---|
| Channel diversification | Reduce single-channel dependency; compare cohort quality, not just CPA |
| CAC optimization | Improve creative, landing, and activation together — cheap clicks that do not activate raise blended CAC |
| Organic vs paid mix | Paid for learning velocity; organic for durability — align budget to PMF stage |
| Virality (k-factor) | Treat K as a sketch; model invites per active user, saturation, and incentive distortion |
Activation optimization
| Lever | Notes |
|---|---|
| Time-to-value | Remove steps between signup and the first successful outcome |
| “Aha moment” | Infer from retention correlates (events that split retained vs churned cohorts) — validate with experiments |
| Onboarding | Checklists, templates, guided setup, and smart defaults |
| Progressive disclosure | Surface depth after the first win — avoid walls of configuration up front |
Activation sequence (conceptual)
Retention optimization
| Topic | Practice |
|---|---|
| Cohort analysis | Group by signup week or campaign; compare curves, not single snapshots |
| Retention curves | Flattening curves suggest PMF in a segment; steep drops flag onboarding or value gaps |
| Engagement loops | Notifications, in-product triggers, and content that tie to recurring jobs |
| Re-engagement | Push, email, and in-app win-back — respect frequency caps and consent |
| Habit (Hook model) | Trigger → action → variable reward → investment — ensure the “investment” stores future value (data, content, workflow) |
Revenue optimization
| Lever | Notes |
|---|---|
| Pricing experiments | Grandfather fairly; test packaging and presentation before destructive list-price wars |
| Upgrade triggers | Usage thresholds, feature gates, and success moments (not arbitrary paywalls) |
| Expansion | Seats, usage tiers, add-ons — align sales and PLG motions |
| LTV / CAC | Targets depend on payback period and capital efficiency; define guardrails (support load, refunds) |
| Paywall UX | Clarity of value, trial design, and payment friction materially affect conversion |
Referral optimization
| Element | Guidance |
|---|---|
| Loop design | Make inviting part of a natural workflow (collaboration, sharing output) |
| Incentives | Two-sided rewards reduce friction; watch fraud and low-quality referrals |
| Tracking | De-duplicate invites; attribute assisted vs direct referral paths |
| NPS | Promoters are a pool for referrals — pair surveys with concrete invite CTAs |
A/B testing infrastructure
| Topic | Guidance |
|---|---|
| Experiment design | Hypothesis, primary metric, guardrails, minimum detectable effect |
| Sample size / power | Pre-calculate before launch; avoid peeking without sequential rules |
| Statistical significance | Frequentist p-values or Bayesian probability — pick one approach per program |
| Sequential testing | Safe interim reads when volume is high |
| Multi-armed bandits | Good for short-lived optimization (headlines, creatives) with clear reward |
| Feature flags | Tie exposure to user/account IDs; support kill switches and gradual rollouts |
| SRM checks | Sample ratio mismatch invalidates many results — monitor allocation drift |
Experimentation culture
| Pillar | Behaviors |
|---|---|
| Velocity | Small batches, clear WIP limits on running experiments |
| Learning backlog | Ideas ranked; “done” includes write-up |
| Team shape | PM + engineer + designer + analyst (or shared analyst pool) for full-stack tests |
| Ethics | No dark patterns; informed consent where required; vulnerable users protected |
Anti-patterns
| Anti-pattern | Consequence |
|---|---|
| Premature growth | Scaling before PMF burns capital and damages reputation |
| Vanity metrics | Optimizing signups without activation or revenue |
| No statistical rigor | False wins and thrashing roadmap |
| Growth hacking without ethics | Regulatory risk, churn, and brand erosion |
External references
- Sean Ellis & Morgan Brown, Hacking Growth — growth process and case patterns
- Reforge — Growth Series and advanced retention / monetization material (paid programs)
- Alistair Croll & Benjamin Yoskovitz, Lean Analytics — metrics, stages, and focus for startups
Index: Growth engineering — index · Marketing map: Marketing body of knowledge
Keep project-specific marketing plans in docs/product/marketing/ and GTM documents in docs/product/, not in this file.