Databricks Analyst Associate Domains Explained

The short version

The exam has nine sections. Querying (20%), dashboards (16%), and query analysis (15%) carry the most weight, while Importing Data and Data Modeling (5% each) are small. Below, each section has what it covers, where candidates slip, and what to practice. Sections follow the Databricks exam guide dated October 30, 2025.

1. Understanding of Databricks Data Intelligence Platform (11%)

What it covers. The platform’s building blocks: Delta Lake for table storage, Unity Catalog for governance, Databricks SQL for analytics, Lakeflow Jobs for scheduled work, and Mosaic AI. You also need the Catalog Explorer: catalogs, schemas, managed and external tables, access controls, views, certified tables, and lineage. Databricks Marketplace is in scope too.

Typical traps. Mixing up what each component owns. Unity Catalog governs access, Delta Lake is a storage format, Databricks SQL is where you query. Another classic: assuming dropping any table deletes its files. Dropping a managed table removes the data; dropping an external table leaves the files in place.

Practice. Click through the Catalog Explorer and find a table’s owner, lineage, and permissions. Create one managed and one external table and drop both.

2. Managing Data (8%)

What it covers. Using Unity Catalog to discover, query, and manage certified datasets, tagging an asset, viewing lineage, and cleaning tables in SQL by removing invalid rows or handling missing values.

Typical traps. `NULL` behavior. `NULL = NULL` is not true, and a `WHERE` filter on a nullable column quietly drops those rows. Know `IS NULL`, `COALESCE`, and `DELETE` with a condition.

Practice. Take a messy table and clean it with SQL, then check the result with counts before and after.

3. Importing Data (5%)

What it covers. The ways data arrives in Databricks: cloud storage such as S3, Delta Sharing with external systems, API-driven intake, Auto Loader, and Marketplace. You also upload a file through the workspace UI.

Typical traps. Choosing the wrong route for the situation. Auto Loader handles new files landing in cloud storage incrementally. Delta Sharing shares live data with another party. A one-off CSV is a job for the upload UI.

Practice. Upload a small CSV to create a table, and read the documentation on Auto Loader so you can explain when you would use it.

4. Executing Queries Using Databricks SQL and SQL Warehouses (20%)

What it covers. The biggest section, with many objectives: using Databricks Assistant in a notebook or the SQL editor, what a SQL warehouse does, joining a Delta table to a federated source, materialized views versus streaming tables versus dynamic views, aggregates (count, approximate distinct count, mean, summary statistics), joins and set operations, sorting and filtering, creating managed and external tables, and Delta time travel.

Typical traps.

`UNION` removes duplicates, `UNION ALL` keeps them.
A `LEFT JOIN` keeps unmatched left rows, and filtering the right table in `WHERE` can turn it into an inner join.
Streaming tables suit append-only, incrementally arriving sources. Materialized views suit precomputed results that are queried often.
Where `NULL` values land when you sort.
A `LOCATION` clause makes a table external.

Practice. Write each join type against the same two tables and compare row counts. Create a materialized view. Query a table `VERSION AS OF` an earlier version.

5. Analyzing Queries (15%)

What it covers. Photon’s features and workloads, finding slow queries with Query Insights and the query profile, auditing a Delta table with its history, using query history and caching, and speeding up filters on large tables with Liquid Clustering. Fixing a broken query is part of this section.

Typical traps. Treating all caches as the same thing. The query result cache, the disk cache, and a materialized view each work differently. Another trap is a time-travel error: if `VACUUM` removed the old files, or the log has aged out, an old version is gone. Liquid Clustering is chosen with `CLUSTER BY`, and it is the modern answer for filter-heavy large tables.

Practice. Run a query twice and compare. Open the query profile and find the slowest step. Read `DESCRIBE HISTORY` output.

6. Dashboards and Visualizations (16%)

What it covers. AI/BI dashboards with multiple pages and datasets and widgets for charts, text, and images; visualizations in notebooks and the SQL editor; parameters in queries and dashboards; sharing with users, groups, and external viewers, plus embedding; scheduled refresh; alerts with a threshold and a destination; and choosing the right chart.

Typical traps. Chart choice questions reward a rule of thumb: distribution is a histogram, relationship between two measures is a scatter, trend over time is a line, parts of a whole is a pie with few slices. A parameter feeds a value into the query, usually in a `WHERE` clause, so know where it is defined and how it is tested.

Practice. Build one dashboard end to end, with a parameter, a schedule, and a shared link. Create an alert that fires on a threshold.

7. AI/BI Genie Spaces (12%)

What it covers. What a Genie space is, how to create one with sample questions, domain-specific instructions, a SQL warehouse, curated Unity Catalog tables and views, and queries vetted as trusted assets. Also permissions, sharing through embedded links, and improving a space by watching questions, accuracy, and feedback.

Typical traps. Permissions have two layers: access to the space, and access to the underlying data in Unity Catalog. Missing either one gives a confusing result. Also, quality comes from curation: clear instructions and trusted queries beat adding more tables.

Practice. Build a space on three tables, ask ten questions, and fix the ones it gets wrong with instructions.

8. Data Modeling with Databricks SQL (5%)

What it covers. Star, snowflake, and data vault models for analytics, and how they line up with the medallion architecture.

Typical traps. Star schemas have a central fact table with denormalized dimensions; snowflake schemas normalize those dimensions further. Bronze holds raw data, silver holds cleaned data, and gold holds data shaped for reporting.

Practice. Sketch a star schema for an orders dataset and place each table in a medallion layer.

9. Securing Data (8%)

What it covers. Unity Catalog roles and sharing settings, the three-level namespace (catalog, schema, table or volume), and good practice for storage and management, including table ownership and protecting personal data.

Typical traps. A user needs privileges at each level: `USE CATALOG`, `USE SCHEMA`, and then `SELECT`. Granting only the last one is not enough. Hiding a column is a column mask, hiding rows is a row filter.

Practice. Grant a test user access step by step and watch what fails until each privilege is in place.

More guides for DATABRICKS-DATA-ANALYST-ASSOCIATE

Next step

Put the weightings to work with the study plan, or try the sample questions. For more practice across all nine sections, see our Databricks Data Analyst Associate practice tests. The vendor’s exam guide is the final authority.