Financial AI often lacks realistic customer context because real household data is sensitive and difficult to use. This dataset takes a different route: it generates fully synthetic Japanese personas shaped by public demographic and financial distributions, so developers can model varied household situations without using real customer records.
What Sets It Apart
- One million records contain nine personas each, for nine million Japanese-language personas across 50 fields.
- Profiles combine age, location, education, and occupation with household structure, income, assets, financial literacy, and money-related concerns. This makes scenarios such as retirement planning, inheritance, and mortgage discussions easier to represent with varied customer contexts.
- Statistical sampling and generated narrative descriptions serve different roles: public distributions ground the attributes, while natural-language profiles support conversational simulation and model evaluation.
- The personas are entirely synthetic and released under CC BY 4.0, enabling commercial and non-commercial use with attribution.
Who It’s For—and the Trade-Off
A good fit for teams developing or testing Japanese-language financial agents, customer simulations, or synthetic training data where privacy constraints rule out real client records. It is not a substitute for observed customer behavior: the profiles reflect available statistics and modeling assumptions, cover adults only, and may not capture every real-world relationship between attributes. Validate suitability for the intended use before relying on simulated outcomes.
