Research Methodology
MotiveGraph develops empirical frameworks for studying demand across markets.
The pages below set out how we separate what is observed from what is inferred, how uncertainty is communicated, how forecasts are meant to be evaluated, and the constraints we place on the use of AI and on sponsored work.
Singapore · Independent market intelligence
Observed behavior
Measured directly.
- recorded transactions
- ownership changes
- retention events
- published prices
Reported with its source, collection window and the transformations applied to it.
Inferred relationships
Estimated, not observed.
- relationships between variables
- estimated effects
- forward-looking estimates
Labelled as estimates. Where a conclusion depends on a model, a comparable set or an assumption, that dependence is stated alongside it.
An estimate that cannot be traced back to observed data is reported as an estimate, not as a finding. We do not claim to know why an individual investor bought, held or sold an asset unless that motive was directly measured.
Observed Data
MotiveGraph distinguishes observed behavior and market data from interpretation. A recorded transaction, an ownership change, a retention event or a published price is data. A statement about why a participant acted is not.
Observed data is reported with its source, its collection window and the transformations applied to it, so that a reader can trace a figure back to what was actually measured.
Inference
Relationships inferred from data are labeled as estimates rather than facts. Where a conclusion depends on a model, a comparable set or an assumption, that dependence is stated alongside the conclusion.
In particular, we do not claim to know why an individual investor bought, held or sold an asset unless that motive was directly measured rather than inferred from behavior.
Model Uncertainty
Quantitative estimates should include assumptions, confidence or uncertainty when appropriate. An estimate presented without its range, sample or assumptions is not useful evidence.
Where uncertainty cannot be quantified reliably, it is described qualitatively rather than replaced with a precise-looking number. A single score that conceals its inputs is treated as a defect, not a feature.
Calibration
Forecasting systems should ultimately be evaluated against realized outcomes.
Where MotiveGraph publishes a forward-looking estimate, the intention is to record that estimate and later compare it with what occurred, and to measure error over time rather than only reporting the cases that resolved favorably.
AI Usage
AI may assist research workflows, but synthetic output should never be presented as observed market data.
Model-generated material may be used for data extraction, classification, research synthesis and hypothesis generation. It is treated as an intermediate step subject to review, not as a source of record, and any published estimate must remain traceable to the underlying data, assumptions and methodology.
Research Independence
Sponsored research, if introduced in the future, should be clearly disclosed and should not determine research conclusions.
MotiveGraph is an independent research project. Where a research direction is supported by an external party, that relationship is expected to be disclosed on the work itself, and the analytical conclusions are expected to remain separate from the sponsorship.
Scope of these statements
This page describes research intent and internal standards. It is not a certification, an audit opinion, or a statement of regulatory status.
MotiveGraph is an independent market intelligence project. Nothing on this page should be read as investment advice or as a guarantee of accuracy.
Questions about methodology are welcome at research@motivegraph.top .