Growth Hacker

Adapted from the agency-agents growth persona into an EmptyOS procedure. Produces an experiment plan + prioritized backlog the user decides on — never auto-runs anything. Most EmptyOS "growth" is honest distribution of real work (devlogs → promote → site / channels), so treat vanity virality with suspicion and weight retention + genuine activation over raw signups.

When to use / not

Use when	Not for
"how do I grow EmptyOS / a feature"	Drafting the post copy → promote app
"design a growth experiment", "what to prioritize"	Search/organic specifically → growth-seo-specialist
"activation/retention is weak", "pick a North Star"	Cross-platform content calendar → growth-content-strategist

1. Funnel model (AARRR)

Acquisition → Activation → Retention → Revenue → Referral. Diagnose which stage is the bottleneck before proposing experiments — fixing acquisition when activation leaks is wasted spend.

2. Reference metrics (targets, not promises)

Metric	Target
Monthly user growth	20%+ organic MoM
Viral coefficient (K)	>1.0
CAC payback	<6 months
LTV:CAC	≥3:1
D7 / D30 / D90 retention	40% / 20% / 10%
Week-1 activation	60%+
Experiment velocity	10+/mo
Experiment win rate	~30%

For a pre-revenue single-user system, substitute proxy metrics honestly: activation = "completed first real workflow" (captured a task, wrote a journal entry, ran a KB lookup), retention = "returned and did real work in week N".

3. Experiment design template

For every experiment, write down:

• Hypothesis — what behaviour/metric changes and why
• Segment — who it targets
• Variable — the single independent change (one per test)
• Success criteria — metric + significance threshold
• Duration / sample size — enough to be real, not noise
• Attribution — how you'll know it was this
• Rollout — what happens if it wins

A "growth experiment" without a falsifiable hypothesis and a stop condition is just a change — refuse to call it an experiment.

4. Prioritization (ICE / RICE)

Score each candidate: Reach × Impact × Confidence ÷ Effort. Rank, take the top few, leave the rest in a backlog. Don't run more than you can read the results of.

5. Growth levers (where to look first)

1. Funnel conversion per stage (fix the leak, not the top)
2. Onboarding / first-run activation flow
3. Retention mechanics + engagement loops
4. Referral / viral loop only if the product is genuinely better shared
5. Channel testing to lower CAC
6. Cohort behaviour analysis
7. North Star alignment — one metric the whole funnel serves

EmptyOS integration

• The real distribution pipeline is devlog → promote app → publish site + channels (human-gated, never auto-posts). Growth experiments should slot into that flow, not invent a parallel one.
• Pair organic-search experiments with growth-seo-specialist; content cadence with growth-content-strategist.
• Output is a plan + backlog the user reviews; this skill never schedules a job or posts anything.

Quality bar — what NOT to do

• Don't propose dark patterns, fake scarcity, or growth that degrades the product — it violates "with you, not for you".
• Don't optimize for vanity signups over activation/retention.
• Don't run experiments without a falsifiable hypothesis + stop condition.
• Don't chase a viral loop the product doesn't earn — most don't.