Business Intelligence and Looker Fundamentals
What the domain tests
The Business Intelligence and Looker Fundamentals domain checks whether you can explain Looker's place in the BI stack. Looker is a semantic layer: LookML projects define models, views, and Explores; the database stores and computes data; the Looker UI saves and presents queries. The exam expects you to distinguish these layers and to know the content objects: Looks, dashboards, folders, and boards. It also tests basic development/production workflow and permissions at a conceptual level.
Core rules
- LookML is version-controlled in Git. Development mode edits a branch; production mode serves the deployed version to users.
- A view maps to a table or derived table and defines dimensions and measures. A model contains Explores and join logic. An Explore is the curated starting point for a query, not a physical table.
- Looker sends SQL to the database; it does not copy the warehouse into its own storage. Persistent derived tables are the exception in that they materialize query results in the database.
- A Look is one saved query. A dashboard arranges multiple tiles, each backed by a query. Folders organize content and carry access settings.
- User roles grant capabilities such as exploring, saving Looks, or editing LookML. Content access is separate: a user must reach the folder and the underlying model.
Worked example
A merchandising manager wants 'revenue by order status for the last quarter, split by region.' The analyst builds an Explore from the orders view, joins customers, selects the status dimension, the region dimension, and the total_revenue measure, adds a date filter, and saves the result as a Look. When the manager opens it, Looker generates SQL with a WHERE clause for the date, groups by status and region, and sums revenue in the database. The LookML defines what total_revenue means; the database does the arithmetic; Looker renders the table. If the manager also needs a KPI tile and a trend, the analyst assembles a dashboard from separate queries rather than assuming one tile feeds another.
Traps
Do not describe an Explore as a database table or a dashboard as a data store. Do not put a measure in a row-level WHERE filter and expect correct pre-aggregation behavior. Do not assume Looker content permissions grant database access. Do not edit in production mode; LookML changes belong in development and must be deployed.
Key points
- Looker is a semantic layer over a database; SQL is generated from LookML.
- Views define fields; models define Explores and joins.
- Looks are single queries; dashboards combine multiple queries as tiles.
- Folders and roles govern content access separately from database rights.
- Development mode and Git control LookML changes; production serves deployed code.