GeoScript Methodology
Numbers that are explainable, versioned, and reproducible.
Every score, metric, and report GeoScript produces can be traced back to real observations. This section explains exactly how — including what the numbers do not measure and where current limitations exist.
How GeoScript approaches methodology
Explainability. Any score GeoScript surfaces can be decomposed into the components that produced it. Scores are not opaque outputs of a machine-learning model — they are deterministic functions of observed inputs with documented weights.
Versioning. When a formula changes, the new version receives a new identifier. Every stored score retains the formula version it was computed under. Historical scores remain valid under their original version; they are not silently recomputed.
Reproducibility. Given the same inputs, GeoScript's scoring functions produce the same output. There is no randomness, no LLM inference, and no Date.now() inside the scoring math.
Honest limitations. GeoScript publishes what it cannot measure alongside what it can. Methodology pages describe data-source gaps, under-evidence behavior, and what a metric is not designed to predict. Limitations are part of the methodology, not an afterthought.
What GeoScript does not do
GeoScript does not claim causal attribution it cannot demonstrate. A change in a score reflects observed inputs changing; it does not automatically prove a specific intervention caused a business outcome.
GeoScript does not treat sponsored placements as organic recommendations. Detection methodology for sponsored content is documented in the Sponsored Detection methodology page (coming soon).
GeoScript does not claim to know hidden model reasoning, paragraph-level attention inside AI engines, or dwell time on a response — unless there is separate, documented evidence for that claim.
GeoScript does not fabricate urgency, manufacture risk scores, or assert purchase probability. The Opportunity Score is not a prediction of whether a business will buy.
Methodology topics
Coming soon
These topics are planned but not yet written. They will be published as the underlying methodology matures and the pages meet GeoScript's documentation quality bar.
- AI Observations
- Prompt Monitoring
- Repeated Observations
- Recommendation Rate
- Share of Voice
- Citation Tracking
- Entity Resolution
- Geo Observations
- Shopping Observations
- Sponsored Detection
- Crawler Verification
- Content Genome
- Outcome Attribution