dbt & Modern Data Stack Vocabulary Exercises

1 / 5

A dbt model contains the expression {{ ref('stg_orders') }}.

What does the ref() function do in dbt?

ref() = dbt's cross-model reference function:

What ref() does	Benefit
Auto-compiles correct schema name per environment	dev/staging/prod schema isolation
Registers model dependency in the DAG	Correct `dbt run` execution order
Powers `dbt docs` lineage graph	Visualise model dependencies

Key vocabulary: model dependency, DAG compile, target schema, ref() vs. source() function.

2 / 5

The CI pipeline reports: "dbt tests failed: not_null on orders.customer_id."

What does this test failure indicate?

dbt tests — data assertions that run against actual data:

Generic test	What it checks
`not_null`	No NULL values in the column
`unique`	No duplicate values
`accepted_values`	Values are in an allowed set
`relationships`	Referential integrity check

A test failure = data quality issue detected in the actual data, not a compile error. Key vocabulary: data test vs. schema test, test severity (warn vs. error), custom test.

3 / 5

A data team says: "The team uses dbt's source freshness check to monitor upstream pipelines."

What does source freshness monitoring in dbt detect?

dbt source freshness — monitoring upstream EL pipeline health:

Configuration	Behaviour
`freshness.warn_after: {count: 12, period: hour}`	Warning if no new data in 12 hours
`freshness.error_after: {count: 24, period: hour}`	Error if no new data in 24 hours

Compares max(loaded_at_field) to current time. Alerts when upstream source has stopped updating. Key vocabulary: source freshness SLO, freshness warning vs. error, loaded_at_field, upstream EL pipeline monitoring.

4 / 5

A team says: "We follow the staging → intermediate → mart dbt project structure."

What is the purpose of the staging layer in dbt?

dbt layered architecture:

Layer	Purpose	Contains
Staging	Clean raw source data	Rename, cast, deduplicate — no business logic
Intermediate	Business logic combinations	Joins, aggregations, derived fields
Mart	Analytics-ready output	Business-facing, stable interface

Key vocabulary: source-aligned model, column renaming, type casting, deduplication in staging, mart as the business-facing model.

5 / 5

A pull request note says: "The dbt project enforces a data contract on the fct_orders model."

What is a data contract in a dbt context?

dbt contracts (dbt 1.5+) — compile-time schema guarantees:

Feature	What it does
`contract: enforced: true`	Enables compile-time contract checking
`columns:` with name/type	Declares the expected output schema
Compile fails on mismatch	Prevents accidental schema changes breaking consumers

Key vocabulary: contract enforced, breaking schema change prevention, downstream consumer stability, column-level contract.