Data Inputs Reference — Helm
Purpose: Normalize raw business inputs, define minimum viable data, and handle missing information without hiding uncertainty.
Contents
- Input tiers
- Minimum input schema
- Recommended and optional inputs
- Industry defaults
- Gap handling
- Handoff-ready input packaging
Input Tiers
| Tier | Meaning | Accuracy impact |
|---|---|---|
Tier 1 |
mandatory minimum | simulation cannot proceed without it |
Tier 2 |
strongly recommended | materially improves accuracy |
Tier 3 |
optional domain-specific input | improves targeted analyses |
Tier 4 |
external defaults | usable with disclosure |
Tier 1: Minimum Input Set
financial_overview:
current_mrr: "¥Xm"
revenue_type: "SaaS|Product|Service|Mixed"
growth_rate_ytd: "X%"
gross_margin: "X%"
operating_profit: "¥Xm or X%"
cash_position: "¥Xm"
monthly_burn: "¥Xm"
primary_model: "SaaS|Transaction|Project|Product|Hybrid"
customer_type: "B2B|B2C|Both"
avg_contract_value: "¥X"
market_context:
industry: "SaaS/HR|EC|FinTech|Healthcare|Manufacturing|..."
target_market: "SMB|Mid-market|Enterprise|Consumer"
geographic_scope: "Japan|APAC|Global"
tam: "¥Xbn"
market_growth_rate: "X%/year"
competitive_position: "Leader|Challenger|Niche|Follower"
simulation_config:
horizon: "SHORT|MID|LONG|ALL"
priority_question: "string"If the user gives a short verbal description, normalize it into this schema before modeling.
Tier 2: Recommended Inputs
KPI Data
kpi_data:
total_customers: X
new_customers_per_month: X
churn_rate_mrr: "X%"
churn_rate_logo: "X%"
nps: XX
cac: "¥X"
ltv: "¥X"
ltv_cac_ratio: "X.X"
payback_period: "X months"
magic_number: "X.X"
arpu: "¥X"
expansion_revenue_rate: "X%"Competitive Intelligence
competitive_intel:
direct_competitors:
- name: "Competitor A"
estimated_revenue: "¥Xm"
market_share: "X%"
key_differentiators: ["...", "..."]
recent_moves: "..."
our_differentiation: "string"
win_rate: "X%"Organization Data
organization:
headcount_total: X
headcount_by_function:
engineering: X
sales: X
marketing: X
cs: X
ga: X
personnel_cost: "¥Xm/month"
personnel_ratio: "X%"
planned_hires_next_12m: X
key_roles: ["...", "..."]Tier 3: Optional Inputs
ESG datafor long-horizon simulationsproduct_portfoliofor BCG and product allocation workma_exit_contextfor M&A or exit analysis
Tier 4: Default External Benchmarks
Use only when the user does not provide data, and always disclose the substitution.
| Input | Default (2026) | Source |
|---|---|---|
| SaaS churn (classical B2B) | 1-2%/month |
Standard benchmark |
| SaaS churn (AI-native, <$250/mo ACV) | 3-7%/month |
m3ter / SaaS Mag 2026 — disclose AI-native flag |
| SaaS gross margin (classical) | 70-80% |
Standard benchmark |
| SaaS gross margin (AI-native) | 60-70% |
SFAI Labs 2026 disclosure tracker; Bessemer Shooting Stars ~60% |
| Japan IT market growth | 3-5%/year |
Standard benchmark |
| CAC payback (B2B SaaS) | 12-18 months |
Standard benchmark |
| LTV:CAC target | 3:1+ |
Standard benchmark |
| SaaS Magic Number healthy line | 0.75+ |
Standard benchmark |
| SaaS M&A revenue multiple | 5-8× ARR (AI premium tier 1-3× higher per Livmo 2026) |
Standard benchmark + AI premium |
| Startup personnel-cost ratio | 60-70% |
Standard benchmark |
| Rule of 40 (public SaaS median) | 28% (Q4 2025), only ~20% clear 40 |
Aventis Advisors 2026 |
| Seed median post-money | $24M (Q4 2025); AI seed +42% vs non-AI |
Carta State of Pre-Seed Q1 2026 |
Gap Handling
Gap Severity
CRITICAL:
- revenue scale unknown
- industry or market unknown
- horizon not specified
SIGNIFICANT:
- churn unknown
- competition unknown
- cost structure unknown
MINOR:
- ESG missing
- M&A details missing
- product portfolio missing
Action Rules
CRITICAL-> triggerON_DATA_INSUFFICIENTand ask firstSIGNIFICANT-> proceed with industry defaults and disclose assumptionsMINOR-> narrow scope and list the limitation in the output
Disclosure Table
| Field | Assumed value | Reason | Replace later |
|---|---|---|---|
| churn | 1.5%/month |
B2B SaaS median | yes |
| gross margin | 75% |
SaaS benchmark | yes |
| market growth | 8%/year |
market benchmark | yes |
Handoff-Ready Packaging
Use the same normalized schema when ingesting COMPETE_TO_HELM or PULSE_TO_HELM. Convert unstructured inputs into Tier-based fields before simulation.