Data & Methodology

1. Data sources

Trend data primarily comes from the production trends_trending_now table (external collector writes Google Trends snapshots). If unavailable, the system falls back to google_trends. init-db inserts demo data only when the table is completely empty and never overwrites existing real data.

2. Opportunity scoring (7-dimension ROI)

Model-defined weights below (not externally cited; disclosed in every business plan):

• marketSignal: 22%
• persistence: 15%
• intent: 10%
• monetization: 14%
• defensibility: 10%
• urgency: 6%
• buildCost: 10%
• timeToRevenue: 8%
• competition: 5%

Three dimensions are data-driven: marketSignal uses search volume and growth (the capped "+1000%" Breakout value is treated as an unquantified-surge sentinel with log dampening, never as a literal percentage); persistence uses each keyword's recurrence statistics across the multi-day collection history (days seen / observations / volume trajectory → flash / recurring / rising); intent uses a transparent rule-based commercial-intent classifier (transactional / commercial / informational / entertainment / ephemeral, with auditable matched signals). Remaining dimensions come from opportunity archetype templates. Low-intent/flash keywords are down-weighted by data, never hard-excluded.

3. Financial model & assumptions

Seed returns use a cash-realized methodology based on public benchmarks (startup survival, VC realized returns, online asset M&A multiples). Exit MOIC probability buckets are model-calibrated assumptions. Auto-BP operating assumptions derive deterministically from archetype dimensions (conversion from monetization, churn from defensibility, CAC from competition) and are clamped to public ranges.

User scale no longer uses one flat constant: year-3 reachable users = the source keyword's peak-day search volume × annualization factor (30) × effective capture share (1% baseline, modulated ±50% by persistence/intent), saturation-compressed into the disclosed [20k, 600k] band — every parameter and its source is in the registry below. LTV/CAC is forced into the credible band (6:1 cap): when raw assumptions imply a higher ratio, the effective CAC is raised to the band edge under a competitive-equilibrium assumption and the adjustment is disclosed in the BP; fantasy ratios like 40:1 are never reported. Average customer lifetime is capped at 3 years (conservative).

4. Auto BP generation

Scheduled cron and API triggers use deterministic templates (no AI): archetype around source keyword, fixed fallback assumptions, code-computed financials, template narrative. Offline npm run bp:generate can use AI for narrative and images. All BPs state they are auto-generated by deterministic models, not manual due diligence.

5. Source registry

1. China startup 1-year survival rate: Caixin, “Enterprise Vitality: A Decade of Chinese SME Insight” (2014–2023 cohorts) (2024-05) · Source link

   Over the past decade, ~92% of newly founded Chinese companies survived their first year.

2. China startup 3-year survival rate: Caixin, “Enterprise Vitality: A Decade of Chinese SME Insight” (2014–2023 cohorts) (2024-05) · Source link

   3-year survival ≈76.0% for 2014–2023 cohorts (annual attrition 8.2% / 9.4% / 6.4%).

3. China startup 5-year survival (interpolated): Interpolated estimate (geometric, between y3 = 0.76 and y10 = 0.503) (2024-05) · Source link

   The report gives no direct 5-year figure; constant-hazard geometric interpolation between years 3 and 10 yields ≈67.5%, explicitly labelled an interpolated estimate.

4. China startup 10-year survival rate: Caixin, “Enterprise Vitality: A Decade of Chinese SME Insight” (2014–2023 cohorts) (2024-05) · Source link

   ≈50.3% of companies survive to year ten.

5. Average Chinese SME lifespan: People’s Bank of China report (widely cited by Chinese media) (2019-06) · Source link

   Average Chinese SME lifespan ≈3 years (US ≈8 years, Japan ≈12 years).

6. Share of VC capital realizing <1x: Correlation Ventures — “Venture Capital, We’re Still Not Normal” (2010s decade (realized)) · Source link

   ≈37% of invested capital realized <1x (a loss); by deal count, roughly half of deals lose money.

7. Share of VC capital realizing ≥10x: Correlation Ventures (2010s decade (realized)) · Source link

   Less than 4% of invested capital realizes ≥10x (the power-law tail).

8. VC return power law: Correlation Ventures — “The 80/20 Rule for U.S. Venture? Not Exactly.” (2010s decade) · Source link

   Returns are highly right-skewed; a small number of winners contribute most of the profits.

9. Exit MOIC distribution (calibrated): Calibration: Correlation Ventures realized-return shape + online-asset M&A multiples (Empire Flippers / FE International / Acquire.com, 2026) (2026) · Source link

   MOIC distribution conditional on a realized cash liquidity event (M&A / secondary / buyback); upside is compressed for small online assets (rarely >25x). Bucket probabilities sum to 1.

10. Annual exit-realization hazard (assumption): Documented assumption: median VC exits take ~5–8 years; small online assets transact faster via Acquire.com / Empire Flippers / FE International; calibrated so the cumulative 5-year exit probability ≈40% conditional on survival. (2026) · Source link

    Cumulative L(t) = 1-(1-h)^t; h = 0.097 → L(5) ≈ 0.40. Explicitly labelled an assumption and stress-tested in the sensitivity analysis.

11. Micro-SaaS ARR multiple: CT Acquisitions / Empire Flippers / Acquire.com market observations (2026) · Source link

    Micro-SaaS (<$1M ARR) typically trades at 2.5–4x ARR.

12. Micro-SaaS SDE multiple: FE International / Empire Flippers (2026) · Source link

    Typically 4–6x seller discretionary earnings (SDE); assets with low owner-dependency fetch the high end.

13. Trend annualization factor (model assumption): Documented model assumption: trending interest decays in pulses; annual topic interest ≈ 30 peak-day equivalents (2026)

    Google Trends volumes are peak-day buckets; annual topic searches ≈ peak-day volume × 30. Explicitly a disclosed model assumption, bounded by the reach limits below.

14. Capture share (model assumption): Documented model assumption: a focused niche site captures ~1% of annual topic search interest at maturity (2026)

    Derived conservatively from SERP click-share distributions (~28% at #1, ~7% at #5, <1% on page 2); modulated ±50% by data-driven persistence/intent scores.

15. Reachable-user bounds (model constraint): Documented model constraint: year-3 reachable users are saturation-compressed into [20k, 600k] (2026)

    Lower bound = minimum viable niche audience; upper bound = realistic single-niche-site capacity ceiling. Applied via a saturating function, not a hard clamp.

16. Zero-human fixed ops base (model assumption): Documented model assumption: hosting/compliance/model-subscription/monitoring base ramps $60k → $90k → $120k over years 1-3 (2026)

    No payroll (zero-human company); includes outsourced legal/finance and exception-handling budget.

17. Per-active-user marginal cost (model assumption): Documented model assumption: ~$0.8 per active user per year for inference + infrastructure (2026)

    Estimated for lightweight AI workflows with caching and batching.

18. USD/CNY exchange rate: Recent approximate CNY-per-USD rate (used for conversion; updated as needed) (2026) · Source link

    Exchange rates fluctuate; converted figures are approximations as of the stated date.

19. Seed-round equity dilution: Industry norm: a single seed round typically dilutes 10%–20% (2026) · Source link

    Baseline 12%; used to convert enterprise-level exit value into the seed investor’s share.

20. Early-stage venture discount rate: Early-stage VC required rates of return are typically 30%–60% (high risk premium) (2010s) · Source link

    Used for risk-adjusted discounting; baseline 35%.

6. Disclaimer

Platform outputs are model projections based on public data with forward-looking assumptions, for research only—not investment, legal, or financial advice. Users must exercise independent judgment.