Metadata
Overview
Metadata in Sift allows organizations to define structured key value pairs that can be applied to any resource in the workspace, creating a consistent taxonomy for organizing, categorizing, and filtering data. Each Metadata key is defined once at the organization level and bound to a specific data type, ensuring values are used consistently across projects while enabling advanced search and filtering to make large collections of resources easier to manage and explore.
Resources that support Metadata
The following Sift resources support user-defined Metadata: Runs, Assets, Annotations, Rules, Reports, Report Templates, Campaigns, Calculated Channels, and User-defined Functions.
Data types for Metadata values
Metadata values support the following data types:
| Type | Description | Example |
|---|---|---|
| String | Text values for categorical data | Production, Test Phase 1, Vehicle_001 |
| Number | Numeric values (integers and decimals) | 1.5, 42, 100.25, -273.15 |
| Boolean | True/false values for binary classification | true, false |
Metadata key properties
The following properties ensure Metadata keys remain consistent and reliable throughout the workspace:
| Property | Description |
|---|---|
| Type immutability | Once a Metadata key is created, its data type cannot be changed. |
| Value consistency | All values for a given key must match the key's defined data type. |
| Organization scope | Metadata keys are shared across the entire organization. |
| Uniqueness | Each Sift resource can only store one Metadata value per key. |
Best practices
The following table summarizes best practices for defining and organizing Metadata in Sift:
| Category | Guideline |
|---|---|
| Naming conventions | Use descriptive, consistent key names (for example, test_phase instead of tp) |
| Follow your organization's naming standards for consistency | |
Consider using prefixes for related keys (for example, vehicle_type, vehicle_ID) | |
| Data organization | Plan your Metadata schema before creating many keys |
Use boolean keys for binary classifications (for example, is_production, validated) | |
Use number keys for quantifiable attributes (for example, version_number, test_count) | |
Keep string values consistent within each key (for example, always use production not prod or PRODUCTION) | |
| Search and filter | Use Metadata to create custom views of your data by filtering objects with specific Metadata values |
| Combine multiple Metadata filters to create precise object collections | |
| Leverage Metadata in Rules and Reports to automatically categorize results |