Keyboard shortcuts

Press ← or → to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

10.5 Evidence-Scored Growth Experiments (High-Variance, Not Default)

Run these only as time-boxed experiments with explicit stop-loss gates. This section is not a default playbook.

Evidence score legend:

  • A (Strong): direct launch-window founder/early-team behavior artifact.
  • B (Moderate): dated launch-period company/media artifact without full unit-economics disclosure.
  • C (Weak transfer): historical control pattern with cross-era/context risk.

Experiment Backlog

  1. Founder-led forum seeding with tracked conversion links (Score: A)

    • Proven in this evidence set: launch-window founder/early-team thread operations and onboarding replies.
    • Unproven in this evidence set: exact conversion lift by thread/community.
    • Stop-loss: stop after 2 weekly cycles if source threads fail funded-activation and D30 quality gates. References: E32 E34 E36 E58 E59 E60

  2. Geo-scoped affiliate + paid stack before expansion (Score: B)

    • Proven in this evidence set: Novig publicly documented affiliate + paid-channel launch support in Colorado.
    • Unproven in this evidence set: partner-level retention quality, fraud-adjusted payback, and durable LTV/CAC.
    • Stop-loss: freeze expansion if retained-trader quality or abuse rates degrade in the pilot geo. References: E65 E66 E69

  3. Referral waitlist flywheel before full-open launch (Score: B)

    • Proven in this evidence set: Robinhood waitlist scale and referral/organic durability (later filing context).
    • Unproven in this evidence set: equivalent effect size in prediction-market onboarding flows.
    • Stop-loss: kill if waitlist growth does not convert into funded first trades with acceptable activation latency. References: E70 E71 E72 E73

  4. Narrow launch inventory to concentrate early liquidity (Score: B/C)

    • Proven in this evidence set: focused early market/contest scope appears in Polymarket and historical DFS controls.
    • Unproven in this evidence set: optimal launch-catalog size for LONGSHOT’s current audience and compliance surface.
    • Stop-loss: expand only when spread/depth/fill and retained cohort quality are stable across two consecutive windows. References: E30 E32 E46 E50 E51

  5. Credibility-first sequencing before aggressive scale (Score: B)

    • Proven in this evidence set: Kalshi’s pre-launch legitimacy + regulated positioning preceded broad public distribution.
    • Unproven in this evidence set: direct causal impact versus alternative distribution sequencing.
    • Stop-loss: if trust/compliance artifacts do not improve activation quality, do not continue funding credibility-heavy campaigns. References: E35 E39 E40

Excluded from Default Backlog

The following patterns were removed as defaults because evidence in this book is too generic or too cross-era for direct transfer:

  • one-line wedge campaign as a universal rule
  • fast-cycle format expansion as a default outside proven launch context

References: E30 E35 E46 E50 E58 E59 E60 E65 E66 E69 E70 E71 E72 E73