Platform/Pillar 7
Private beta

Data

Every observation, recommendation, and outcome lives in a structured, tenant-isolated, immutable-where-it-matters data architecture — the evidence base the entire platform reasons over.

ReportingEnd of loop — feeds back to start →

Four data planes

GeoScript's evidence base is organized around four data planes: the observation plane (raw AI engine responses, preserved before parsing), the content plane (the Content Genome — structured extraction from a client's website), the deployment plane (signed intervention packages, preflight records, verification results), and the measurement plane (before-and-after visibility attribution, report snapshots). These planes are kept separate — each is immutable where it matters and feeds the next stage in the product loop.

Tenant isolation: the foundational rule

Every data record in GeoScript is owned by a specific agency workspace. The platform enforces tenant isolation at the service layer, not just the query layer — a workspace cannot access another workspace's observations, clients, opportunities, or reports. This is enforced at the database policy level and is treated as a Tier 1 invariant: code that violates tenant isolation does not ship.

Immutability where it matters

AI visibility observations are immutable — once recorded, a raw response cannot be altered. Report snapshots are immutable — once published, the numbers are frozen. Opportunity score components are versioned — formula changes produce new version records, not overwrites. This design means the data underlying a report or pitch room accurately reflects what was true at the time it was recorded, not a retroactively adjusted version.

Determinism: the observable behavior standard

GeoScript's data processes are deterministic wherever possible — the same inputs produce the same outputs. The prompt universe generator produces the same 1,118 prompts for the same client inputs every time. The Content Genome extractor produces the same typed blocks from the same page content every time. The Optimization Engine produces the same recommendations from the same genome every time. This makes the system auditable: unexpected outputs point to changed inputs, not random behavior.

Methodology transparency

GeoScript publishes the methodology behind its scored metrics and data processes. The Opportunity Score methodology documents every component, its weight, the missing-evidence penalty, and the formula version. The data freshness methodology documents how timestamps are assigned, how stale evidence is handled, and why published report snapshots are isolated from live data updates. The methodology section is kept current as the platform evolves.

What Data is not

Explicit limitations (§127)

  • GeoScript does not sell, license, or share client observation data with third parties.
  • GeoScript does not use one client's data to benefit another client's recommendations or scores.
  • GeoScript does not claim perfect causal attribution — the measurement plane records before-and-after differences; causal inference requires acknowledging confounders.
  • GeoScript does not store or process personally identifiable information about a client's end-customers — observations are about AI engine responses, not individual consumer data.
  • GeoScript is not a general-purpose data warehouse or analytics platform — the data architecture is purpose-built for the AI visibility measurement and optimization loop.

Related

Continue through the loop

Data is pillar 7 of 7.

Other pillars