Friendly Names: Add human-readable labels at varying levels of verbosity. These labels are carried through to model logic, insights, and explanation layers.

Units and Dimensions: Define units (e.g. days, dollars) and dimensions (e.g. time, volume) to improve interpretability and support dimensional reasoning.

Protected Features: Flag attributes such as age, gender, ethnicity, or other legally or ethically sensitive variables. The platform uses this tagging to enable fairness checks, bias mitigation strategies, and compliance with regulatory frameworks (e.g. GDPR, EEOC, EU AI Act).

Feature Groups: Organise features into logical groupings to support structured modelling and explanation across related variables.

Feature Value Groups: Cluster specific values within a feature (e.g. product categories, risk tiers) to enable abstraction and simplify reasoning. API Guide >>

High Cardinality Treatments: Apply compression, grouping, or encoding strategies to manage features with large, sparse value sets.

Missing Value Treatments: Set rules or preferences for imputing, excluding, or flagging missing values during training and inference.

Value Types and Roles: Specify whether features represent identifiers, flags, continuous measures, temporal data, or reference variables. This helps the platform apply appropriate transformations and reasoning patterns.

Causal Tags: Annotate features with causal roles (e.g. predictor, mediator, outcome) to support causal inference and scenario simulation.

Data Provenance: Track the source and lineage of each feature to support traceability and compliance.

Policy Constraints: Define limits, thresholds, or guardrails that reflect business rules or regulatory requirements. These can be carried into model logic and validation.

Temporal Relevance: Mark features with time sensitivity or decay characteristics (e.g. last updated, expiry) to support behavioural reasoning and time-based partitioning.

Default Treatments: Set default behaviour for unseen values, rare categories, or ambiguous cases during inference.

Semantic Types / Ontology Links: Map features to standard taxonomies or domain ontologies to support semantic consistency and integration across models.