Glossary

The terms used throughout these docs and in the files. The source is GLOSSARY.md at the repo root.

Data products

Data product: A named, owned unit of data on the platform that reads through input ports and publishes through output ports. Avoid: Product, dataset, project

Owner: The person accountable for a data product and the one to call when a number is wrong. Avoid: Business owner, steward

Technical owner: The person who builds and maintains a data product. Avoid: Maintainer, developer

Data product kind: A label on a data product taken from a list the platform defines, by default source-aligned, aggregate or consumer-aligned. Each kind has a short prefix used in derived names. Avoid: Product type, layer, tier

Naming template: A platform-wide pattern that produces a derived name, such as a port name or a role name, from a data product’s kind and name. Avoid: Naming rule, pattern

Domain: The business area a data product belongs to, such as finance, sales or operations. Avoid: Department, area

Ports

Port: A named interface of a data product through which data enters or leaves. Its name comes from a naming template and is unique across the platform. Avoid: Interface, endpoint, dataset

Private schema: The place where a data product lands its ingested tables and builds its output tables. Nobody outside the data product reads it. Avoid: Staging schema, internal schema

Input port: A port through which a data product receives data, either from a source or from another data product’s output port. Avoid: Dependency, upstream

Output port: A port through which a data product publishes data, with its own freshness promise and access policy. It exposes views only, never the tables themselves. Avoid: Public schema, export

Default output port: The output port that carries the same name as its data product.

Port kind: The form a port takes: a schema holding several tables, or a single table. Other forms may follow. Avoid: Port type

Table: One table in a port, with the columns and quality rules expected of it. Avoid: Model, dataset

Quality rule: A condition the data in a column or a table must meet, such as not null, unique or a row count bound. Avoid: Test, check, expectation

Severity: What happens when a quality rule is broken: error fails the run, warning reports and continues, none ignores the rule. Avoid: Level, priority

Source: A system outside the platform that delivers data into an input port. Avoid: Source system, upstream system

Promises on an output port

Reader: A group, role, user or data product that is allowed to read an output port. Avoid: Consumer, subscriber

PII reader: A reader that may also see the columns marked as personal data.

Freshness: The promise of how old the newest data in an output port may be, stated as a duration. Avoid: SLA, latency

Building tables

Contract: The part of a table’s definition that states what to expect from it and names no technology. Avoid: Spec, schema definition

Build: The part of a table’s definition that states how the table is made: its executor, its code and the parameters passed to it. Avoid: Runner, pipeline config

Executor: The technology that builds a table, such as Python or dbt. Avoid: Runner, engine

Orchestrator: The system that triggers builds on a schedule, such as Airflow. The platform usually has one. Avoid: Scheduler

Runtime: The place where an executor’s code runs, such as an Airflow worker or a container. Avoid: Environment, compute