SKILL.md
Growth Strategy
Modern growth hacking: loops + product-led growth + disciplined experimentation, under privacy and deliverability constraints.
When to Use
- Designing growth strategy or GTM plans
- Planning experiments and A/B tests
- Optimizing activation, retention, or referral flows
- Building viral/referral loops
- Reviewing growth tactics for ethics/compliance
Core Principle
If a "hack" doesn't strengthen a loop or an input metric, it's noise.
1. Growth Model First
North Star Metric (NSM)
- Single metric aligning the whole org
- Plus input metrics (leading indicators you can move weekly)
- Avoid vanity metrics
Growth Loops > Funnels
- Loops: Closed systems where outputs feed inputs → compounding growth
- Funnels: Linear → diminishing returns
Common loops:
| Loop Type | Example |
|---|---|
| Viral | User creates → shares → new users |
| UGC/SEO | User creates content → indexed → new users find |
| Paid | Revenue → reinvest in ads → more revenue |
| Sales | Customer → case study → new leads |
Product-Led Growth (B2B/SaaS)
Product itself drives: Acquisition → Activation → Retention → Monetization
2. Instrumentation
Event Taxonomy
- Clean identity resolution: anonymous → user → account
- Cohort retention tracking
- Activation milestones defined
Incrementality
- Holdouts / geo splits when attribution is noisy
- Don't trust last-click blindly
Metric Categories
| Type | Examples |
|---|---|
| Core | NSM + input metrics |
| Guardrails | Churn, spam rate, refunds, latency, NPS |
3. Experimentation Engine
Intake System
- Single queue + scoring (RICE/ICE)
- Weekly cadence
Test Definition (Required)
- Hypothesis
- Target segment
- Success metric
- Guardrail metrics
- Sample size rule
- Kill criteria
High-ROI Test Areas
- Onboarding steps
- Paywall copy
- Pricing/packaging
- Referral incentive
- Landing page variants
- Lifecycle messages
4. Lever-Specific Playbooks
Activation & Onboarding (Highest ROI)
- Reduce time-to-value
- Templates, importers, "one-click first win"
- Progressive disclosure (ask when needed, not upfront)
- Guided setup flows
Viral/Referral Loops
- Build shareable artifacts (reports, badges, embeds)
- "Invite teammates" as natural workflow
- Reward activated referrals, not just signups
Content + SEO
- Programmatic SEO: template + real value + strong linking
- Audit/prune thin pages (don't endlessly generate)
- Quality > quantity
Lifecycle (Email/Push)
Deliverability is gating factor:
- SPF/DKIM for all senders
- DMARC for bulk
- Keep complaint/spam rates low
Community-Led Growth
- Seed right early members
- Great "first experience"
- Connect to business outcomes (support deflection, referrals)
5. Privacy & Measurement Constraints
Expect
- Less reliable cross-site tracking
- Cookie-based attribution unstable
- Platform policy changes
Adapt
- First-party data focus
- Server-side signals
- Incrementality testing
- Design measurement that survives policy changes
6. AI in Growth
Good Uses
- Generate creative/landing page variants to test (humans review)
- Summarize qualitative feedback
- Cluster objections
- Speed up research
Avoid
- "AI content spam" at scale without quality control
- Backfires in SEO and brand
7. Hard Red Lines
If a tactic can't survive being in a postmortem or public doc, don't ship it.
Never:
- Spam (email/SMS)
- Fake reviews
- Scraping that violates ToS
- Dark patterns
- Deceptive pricing/consent
Output Format
When proposing growth initiatives:
## Initiative: [Name]
**Loop/Lever**: [Which growth loop or lever this strengthens]
**Hypothesis**: [If we do X, Y metric will improve by Z because...]
**Input Metric**: [What leading indicator we're moving]
**Guardrails**: [Metrics that must not regress]
### Implementation
[Concrete steps]
### Measurement
[How we'll know it worked]
### Kill Criteria
[When to stop if failing]Quick Checklist
Before shipping any growth tactic:
- Does it strengthen a loop or input metric?
- Is the hypothesis testable?
- Are guardrails defined?
- Is it compliant with platform ToS?
- Would you put it in a public doc?
- Does it respect user privacy?
- Is deliverability accounted for (if email)?
See references/ for detailed playbooks.