Materialisations for Analytics Engineers
Materialisations decide how dbt builds a model in the warehouse: as a view, a table, an incremental table or an ephemeral CTE. The choice balances build time, storage and freshness, and incremental strategies such as merge, delete and insert, or insert overwrite control how new rows are applied.
Stage: dbt in Depth · Part of the Analytics Engineer roadmap · Last reviewed 19 November 2026
What to learn in Materialisations
- Choose between view, table and incremental for a given model
- Write an incremental model with an is_incremental filter
- Pick a unique key and an incremental strategy
- Run a full refresh when logic changes
- Handle late-arriving rows in an incremental model
Materialisations interview questions
When is an incremental model a good choice?
When a table is large and new data arrives as appended or updated rows, so rebuilding it fully each run is slow or costly. Incremental models process only new or changed rows, though they add logic that needs careful testing.
Why does an incremental model need a unique key?
With a unique key, dbt can update existing rows rather than insert duplicates when a record changes. Without it, reprocessed rows would be appended again and inflate counts.
Where Materialisations fits in the Analytics Engineer roadmap
Materialisations belongs to the dbt in Depth stage (intermediate level), which takes about 7 hours. The core craft.
Other units in this stage
Back to the full Analytics Engineer roadmap