Data Products

A contract describes one dataset. A data product is a useful offering with a defined purpose and an accountable owner, usually published as several contracted outputs — tables, views, APIs or semantic models.

OLC keeps the two apart:

• the product is defined once, in its domain's _domain.yaml (products:);
• each contract references the product that publishes it (info.data_product), by a stable id.

Defining a product

# _domain.yaml
domain: real_estate

products:
  - id: property_performance          # required · lowercase id, set once, never changed
    name: Property Performance        # required · display name, safe to rename
    description: Monthly property income, expenses and operating performance.
    owner: property_data_team         # optional · falls back to the domain's ownership
    lifecycle: active                 # proposed (default) | active | deprecated | retired
    expected_outputs:                 # optional · declare when completeness matters
      - id: monthly_property_income   # required · unique within the product
        name: Monthly Property Income
        kind: table                   # table (default) | view | api | semantic_model
        required: true                # default true
      - id: occupancy_model
        name: Occupancy semantic model
        kind: semantic_model
        required: false

Field	Rules
id	^[a-z][a-z0-9_]{1,62}$. Unique within the domain. The value contracts reference — never change it.
name	Required. Change freely; no contract refers to it.
lifecycle	proposed, active, deprecated, retired. Defaults to proposed.
expected_outputs[].id	Same id format. Unique within the product.
expected_outputs[].kind	table, view, api, semantic_model.
expected_outputs[].required	Whether the product is incomplete without it.

Unknown keys are rejected by strict validation, except x- vendor extensions. An empty products: means none.

Referencing a product from a contract

info:
  title: gold_property_monthly_income
  domain: real_estate
  data_product: property_performance            # the product that publishes this output
  data_product_output: monthly_property_income  # optional · which expected output this implements

• info.data_product means the product that publishes this contracted output — not every product that uses it.
• Both values are ids in the same format as above. data_product_output requires data_product.
• A contract without data_product is valid. Supporting and unassigned datasets are legitimate; whether a contract must belong to a product is a policy your platform applies, not a schema rule.

Shared datasets

A dataset belongs to the product that publishes it, or to none. When another product depends on it — a shared dimension, a common silver table — that is a lineage relationship (upstream_contracts), not a second membership.

Environments

A product is defined once and means the same thing in every environment. Whether an output exists, and what evidence it reports, varies by environment — that is observed, not declared.

What a validator checks, and what it cannot

Strict validation checks each document on its own: id formats, allowed values, unique ids, unknown keys. Checks that need the whole registry — a contract referencing a product id no domain defines, a required output no contract implements, an output claimed by two contracts — belong to the platform that reads all documents together.