Question before model

The workflow begins by stating the decision problem, what evidence could change the view, and which uncertainties cannot yet be resolved. This prevents the system from starting with a preferred model and searching for justification.

Evidence chain

Sources, data transformations, statistical tests and model outputs are linked so that a later reviewer can reconstruct how the conclusion was formed.

• Question
• Source and data
• Statistics
• Quantitative challenge
• Strategic context
• Portfolio and risk
• Governed decision record

Human judgment

The architecture is designed to structure judgment, not to pretend judgment disappears. Conflicting evidence, model uncertainty and mandate constraints remain visible at the decision stage.

Product role

Saptanga is intended to carry these workflows as software. Mission Saptanga remains the institution defining the research and decision architecture.

Methodology

• Start with a falsifiable research question or explicit decision problem.
• Attach source provenance and distinguish observation from inference.
• Use statistical and quantitative challenge where the question supports it.
• Add macro, portfolio and risk context before a decision record is considered complete.

Evidence requirements

• Source trail appropriate to the research question.
• Versioned calculations or model outputs where quantitative methods are used.
• Explicit assumptions and unresolved uncertainties.
• Decision-state and review-state record.

Limitations

• Not every research question requires every engine or system.
• More computation does not compensate for weak source quality or an undefined question.
• Public research methodology is informational and is not a substitute for a regulated personalized engagement.

Connected research architecture

• research
• methodology
• computational architecture
• platform