Growth Marketer

# CLAUDE.md — Growth Marketer
> Version: 2026-02-17 | Difficulty: advanced

## What Your AI Can Now Do
- Analyze your full funnel from acquisition to retention and pinpoint exactly where users drop off
- Design and document A/B experiments with hypothesis, variants, and success criteria in one command
- Pull channel performance data and calculate true CAC and LTV across every acquisition source
- Generate landing page copy variants optimized for specific audience segments and conversion goals
- Build weekly growth reports that track the metrics your team actually cares about

---

## Quick Start
1. Paste this file into your project as `CLAUDE.md`
2. Add your analytics, ad platform, and email tool credentials to `.env`
3. Run `/funnel-analysis` to map your current conversion funnel and find the biggest drop-off
4. Run `/experiment-design [hypothesis]` to structure your first A/B test
5. Run `/channel-report` to see which acquisition channels are actually profitable

---

## About Me
Growth marketer owning the full acquisition and activation funnel. I run experiments across paid, organic, email, and product-led channels. I manage $10K-100K+ monthly ad spend, design landing pages, build email sequences, and obsess over conversion rates. What matters most: sustainable, profitable growth — not vanity metrics or one-time spikes.

---

## Pre-Built Commands

| Command | What It Does |
|---------|-------------|
| `/funnel-analysis` | Maps your conversion funnel stage by stage, calculates drop-off rates, and highlights the biggest leverage point |
| `/experiment-design [hypothesis]` | Structures an A/B test: hypothesis, control, variant, sample size, duration, and success metric |
| `/channel-report [period]` | Pulls spend, conversions, CAC, and ROAS per acquisition channel for the given period |
| `/landing-page [audience] [goal]` | Generates headline variants, body copy, and CTA options optimized for a specific segment and conversion goal |
| `/cohort-analysis [timeframe]` | Breaks users into cohorts by signup date and tracks retention, activation, and revenue over time |
| `/growth-model [assumptions]` | Builds a growth projection model with user acquisition rate, churn, and revenue per user inputs |

---

## How Claude Connects to Your Tools

| Tool | Why You Need It |
|------|----------------|
| Google Analytics | Track website traffic, conversion events, user behavior flows, and attribution across all channels |
| AnySite | Analyze competitor landing pages, scrape ad copy, and research market positioning for new channels |
| Google Workspace | Store experiment docs, share growth reports, coordinate with design and engineering on test implementations |
| Resend | Send and track email sequences for activation, onboarding, and re-engagement campaigns |

### Setup
```bash
# Required in .env
GOOGLE_ANALYTICS_PROPERTY_ID=your-ga4-property-id
GOOGLE_CREDENTIALS_PATH=./credentials.json
RESEND_API_KEY=your-resend-api-key
AD_PLATFORM_API_KEY=your-ad-platform-key
```

---

## Growth Rules
- **MUST:** Never scale a channel before proving unit economics — a cheap CAC means nothing if LTV does not cover it
- **MUST:** Every experiment needs a written hypothesis, success metric, and minimum sample size before launch
- **MUST:** Kill losing experiments fast — if results are statistically significant and negative, stop spending immediately
- **SHOULD:** Test one variable at a time — multivariate tests need 10x the traffic to reach significance
- **MUST:** Track cohort retention, not just top-of-funnel acquisition — growing users who churn in week 2 is burning money
- **SHOULD:** Retest winning experiments quarterly — what worked 6 months ago may not work with a different user mix
- **SHOULD:** Attribute conversions across the full journey, not just last-click — multi-touch attribution reveals the real story
- **MUST:** Never report on metrics without context — "signups are up 20%" means nothing without knowing CAC, activation rate, and retention

---

## Project Structure
```
growth/
├── experiments/
│   ├── active/
│   ├── completed/
│   └── backlog/
├── funnels/
│   └── [funnel-name]/
├── channels/
│   ├── paid/
│   ├── organic/
│   ├── email/
│   └── referral/
├── landing-pages/
│   └── [campaign-name]/
├── analytics/
│   ├── weekly-reports/
│   └── cohort-analyses/
└── models/
    └── growth-projections/
```

### What These Folders Mean
- **experiments/** — A/B tests organized by status: active (running now), completed (with results), and backlog (ideas to test next)
- **funnels/** — Conversion funnel maps with stage-by-stage metrics and drop-off analysis
- **channels/** — Performance data and strategy docs per acquisition channel
- **landing-pages/** — Copy variants, design briefs, and test results for campaign landing pages
- **analytics/** — Weekly growth reports and cohort retention analyses
- **models/** — Growth projection spreadsheets with scenario modeling for planning and fundraising

---

## Automation

| When | What Happens | If It Fails |
|------|-------------|-------------|
| Monday morning | Generates weekly growth report: signups, activation, retention, revenue, and channel breakdown | Pull last available data snapshot and flag which metrics are stale |
| Experiment reaches sample size | Calculates statistical significance and recommends: scale, iterate, or kill | Alert you that the experiment needs manual analysis and pause ad spend if applicable |
| New landing page published | Runs a conversion checklist: headline clarity, CTA visibility, page speed, mobile responsiveness | Log the page as "unaudited" and add it to the next weekly review |
| Monthly (1st) | Recalculates CAC and LTV by channel and flags any channel where CAC exceeds LTV | Use last month's numbers and highlight that they are outdated |
| Ad spend exceeds daily budget by 20% | Alerts you immediately with the channel, current spend, and projected daily total | Pause the campaign automatically if possible, otherwise send an urgent alert |

---

## Gotchas
- **MUST:** Never declare a test winner before reaching statistical significance — early results are noise, not signal
- **MUST:** Never optimize for a proxy metric without validating it correlates with revenue — high signup rates with zero activation is a vanity trap
- **SHOULD:** Document every experiment result, including failures — failed tests teach more than successful ones
- **MUST:** Never scale ad spend more than 20% per day on a winning campaign — algorithms need time to adjust, and CPAs spike on sudden budget jumps
- **SHOULD:** Always check for seasonality before attributing a traffic change to your latest experiment — correlation is not causation

---

## Memory & Learning
Claude builds context over time:
- Which experiments won, lost, and by what margin — so you never repeat a failed test
- Channel-level CAC and LTV trends over quarters, not just snapshots
- Landing page copy patterns that consistently convert for your audience
- Seasonal traffic and conversion patterns that affect experiment timing
- Your funnel benchmarks so Claude can flag when something drifts from normal

To save something important: just say "remember that [fact]" in conversation.
To review saved memories: use the `/memory` command to open the memory editor.