Agent-based market adoption simulator -- Simulate how your service idea would spread in the market.
Place autonomous agents in a virtual representation of the Japanese market and run agent-based simulations powered by Bass diffusion theory. Visualize adoption curves, revenue projections, and improvement opportunities.
Enter a description of your service idea in the text area. Click "AI Analysis" and Claude API will automatically infer simulation parameters (JTBD, target users, category, pricing, competition).
The AI-inferred JTBD (jobs), target demographics, category, suggested price, and competition level are shown in the left panel. Parameters can be manually adjusted. Simulation results are displayed in real-time on the dashboard.
| Metric | Description |
|---|---|
| Estimated TAM | Total Addressable Market size in Japan, estimated by AI based on service description |
| Total Adopters | Cumulative number of agents who adopted the service during the simulation period |
| Peak Daily Adoption | Highest number of new adoptions in a single day. Indicates the peak momentum |
| Adoption Rate | Percentage of TAM that adopted the service. Measures market penetration |
| Market Opportunity Score | ODI (Outcome-Driven Innovation) score (0-10). Higher = more unmet needs in the market = higher chance of success |
What Claude API infers from your description:
- Jobs (JTBD): What users are trying to accomplish -- Functional, Emotional, and Social jobs
- Target: Age groups (20s, 30s...), occupations (office worker, student...), household types (single, couple...)
- Category: Best match from 25 categories (education, healthcare, SaaS, etc.)
- Suggested Price: Recommended monthly price based on market benchmarks and target willingness to pay
- Competition: None / Weak / Strong -- assessment of competitive landscape
- Reasoning: Explanation of the AI's inference process, including TAM estimation rationale
- Blue solid line (Cumulative): S-curve (sigmoid). Shows the progression from innovators through early adopters to majority
- Green dashed line (Daily): Bell curve. New daily adoptions. The classic Bass diffusion pattern: rise, peak, then decline
- Red line (Current position): Position of the timeline slider
Timeline controls: Drag the slider to any simulation day. Press play for animated playback. Adjust speed (0.5x / 1x / 2x / 4x).
Horizontal bar chart showing agent counts at each stage of the AIDMA purchase funnel:
| Stage | Description |
|---|---|
| Unaware | Agents who don't know the service exists |
| Aware | Agents who know about the service but aren't interested |
| Interest | Agents actively gathering information |
| Consideration | Agents specifically considering purchase |
| Adopted | Agents who have purchased/adopted the service |
Transition probabilities between stages are determined by: Bass diffusion (innovation coefficient p + imitation coefficient q), JTBD fit score, price acceptance, and social network influence from neighboring agents.
- Agent-based market simulation -- Generate agents based on Japan demographics (e-Stat 2020 census). Connect via Watts-Strogatz small-world network. Individualize behavior by Rogers adoption categories (Innovators → Laggards)
- Bass diffusion + AIDMA funnel -- Combine innovation/imitation coefficients with a 5-stage purchase funnel model
- JTBD + ODI opportunity score -- 25 category presets with ODI scoring (
importance + max(importance - satisfaction, 0)) - Claude API auto-inference -- Just describe your service; AI infers JTBD, target, category, pricing, and competition
- What-If analysis -- Instantly preview the impact of parameter changes
git clone https://github.com/PM-DAO/value-simulator.git
cd value-simulator
docker compose up
# Open http://localhost:3000Prerequisites: Python 3.11+, Node.js 20+, uv
# Backend
cd backend
uv sync
PYTHONPATH=src uv run uvicorn simulator.api:app --reload # http://localhost:8000
# Frontend (separate terminal)
cd frontend
npm install
NEXT_PUBLIC_API_URL=http://localhost:8000 npm run dev # http://localhost:3000Frontend (Next.js 16) Backend (FastAPI)
┌──────────────────┐ ┌──────────────────────────────┐
│ Input Form │ REST │ api.py ← Endpoints │
│ Dashboard │ ──────→ │ engine.py ← Simulation │
│ Scenario Compare│ JSON │ agent.py ← BDI Agents │
│ (Recharts) │ │ population.py ← Demographics │
└──────────────────┘ │ network.py ← Social Graph │
│ diffusion.py ← Bass Model │
│ funnel.py ← AIDMA Funnel │
│ jtbd.py ← JTBD/ODI + LLM │
│ market.py ← Market Engine │
└──────────────────────────────┘
| Module | Description |
|---|---|
agent.py |
BDI (Beliefs-Desires-Intentions) agent model with demographics, Rogers category, funnel stage |
population.py |
Stratified sampling based on Japan census data (age, gender, income, region) |
network.py |
Watts-Strogatz small-world network (k=6, p=0.1) for social graph construction |
diffusion.py |
Bass diffusion: innovation coefficient (p) for spontaneous adoption, imitation coefficient (q) for word-of-mouth |
funnel.py |
AIDMA purchase funnel with stage transitions and decay (drop-off) at each stage |
jtbd.py |
JTBD/ODI opportunity scoring + Claude API LLM inference for auto-parameter detection |
market.py |
Price model application (free/freemium/subscription/usage/one-time), competition effects |
engine.py |
Simulation orchestrator. Runs all modules step-by-step (daily) with marketing event injection |
api.py |
FastAPI endpoints for simulation, auto-inference, and what-if analysis |
| Endpoint | Description |
|---|---|
POST /api/simulate |
Run simulation with manual parameters |
POST /api/simulate/auto |
AI auto-inference mode (Claude API infers parameters) |
POST /api/simulate/compare |
Scenario comparison (up to 3 simultaneous scenarios) |
API docs (Swagger UI): http://localhost:8000/docs
To use AI auto-inference (automatic JTBD/target/pricing detection from service description), set your Anthropic API key:
export ANTHROPIC_API_KEY=sk-ant-...The simulator works without an API key -- you can manually set parameters and run simulations.
# Backend unit tests
cd backend && uv run pytest
# With coverage
cd backend && uv run pytest --cov
# E2E tests (auto-starts backend + frontend)
cd frontend && npx playwright test| Layer | Technology |
|---|---|
| Backend | Python 3.11+, FastAPI, NetworkX, NumPy, SciPy, Pydantic |
| Frontend | Next.js 16 (App Router), React 19, Tailwind CSS v4, Recharts |
| LLM | Anthropic Claude API |
| Test | pytest, Playwright |
| Infra | Docker Compose |
Development happens in a private repository. Stable releases are synced to this public repository.
Private repo (daily dev) Public repo (this repo)
├── Daily development ├── Release versions only
├── WIP / experiments ├── Clean commit history
└── Testing └── Tagged versions
│
└── When stable ──squash push──→
- Self-host: Clone this repo and run it as-is
- Code audit: All source code is available for review
- Contribute: Submit PRs to this repository
Pull Requests are welcome.
- Fork this repository
- Create a feature branch (
git checkout -b feature/amazing-feature) - Commit your changes (
git commit -m 'Add amazing feature') - Push and create a Pull Request
Bug reports and feature requests: Issues
This software is licensed under AGPL-3.0. If you provide this software as a SaaS (network service), you must make the source code available to users accessing it over the network. See LICENSE for details.