Data Engineer Associate · Databricks
Databricks Data Engineer Associate: read this before the sample questions
The Databricks exam guide comes with sample questions, but there is a catch to know before you rely on them. Plus the seven sections, their weights, and the objectives that ask you to choose.
The Databricks Certified Data Engineer Associate exam assesses whether you can use the Databricks Data Intelligence Platform for foundational data engineering: ingesting, loading, transforming and modelling data (including ETL with PySpark), orchestrating it with Lakeflow Jobs, shipping it with CI/CD, and keeping it observable, fast and governed.
Databricks sets no prerequisite, but recommends course attendance and six months of hands-on experience with the platform.
What the exam measures
| Section | Weight |
|---|---|
| 1. Databricks Intelligence Platform | 6% |
| 2. Data Ingestion and Loading | 21% |
| 3. Data Transformation and Modeling | 22% |
| 4. Working with Lakeflow Jobs | 16% |
| 5. Implementing CI/CD | 10% |
| 6. Troubleshooting, Monitoring, and Optimization | 10% |
| 7. Governance and Security | 15% |
Ingestion and transformation together are 43% of the exam. The platform section, at 6%, is the smallest, but the operational sections (jobs, CI/CD, troubleshooting) add up to more than a third.
Which version you will sit
The exam guide says it covers the exam version as of 4 May 2026, and that it will be updated whenever the exam changes and when those changes take effect.
What the guide does not have is a revision history. Databricks does not publish what changed from the previous version, so the May 2026 guide is the only reference point. If your material predates it, compare it section by section against the current outline rather than relying on anyone's summary of the differences.
The sample questions at the end of the guide come from a previous version of the exam. They show the style, not necessarily the current scope.
How the exam is scored and delivered
- Format: 45 scored multiple-choice questions in 90 minutes. No test aids are allowed.
- Unscored items: the exam may include unscored items, not identified on the form, with extra time allowed for them.
- Delivery: proctored, online or at a test center.
- Languages: English, Japanese, Portuguese (Brazil) and Korean.
- Validity: 2 years. To recertify, you take the current live exam again.
Objectives that ask you to choose
- Ingestion pattern. Batch, streaming or incremental; COPY INTO, Auto Loader (with schema enforcement and evolution), Lakeflow Connect standard or managed connectors, or JDBC/ODBC and REST clients from a notebook. The guide asks you to prioritize between them based on requirements.
- Job triggers. Scheduled, file arrival or table update, and the choice between time-based and data-driven triggers depending on when the data is available and what depends on it.
- Managed or external tables. Differentiate them in Unity Catalog and perform basic operations, including converting between the two.
- Compute. The platform section is small, but it asks you to select the compute option that fits a workload, knowing each one's characteristics, limitations and cost model.
- Where a bottleneck is. Data skew, shuffling and disk spilling, identified from stage-level metrics in the Spark UI, next to the Lakeflow Jobs run history and task graph for pipeline health.
A study plan that follows the guide
- Transformation (22%) and ingestion (21%). Cleaning bronze data with PySpark and SQL, column and row manipulation, joins (inner, left, broadcast, multiple keys, cross) and unions, data quality rules for silver and gold tables, and every ingestion route listed above.
- Lakeflow Jobs (16%). Control flow with retries, branching and looping; notebook, SQL query, dashboard and pipeline tasks with their dependencies; schedules and triggers.
- Governance and security (15%). Table types, GRANT, REVOKE and DENY for users, groups and service principals, column masking and row-level security, and Unity Catalog ABAC policies.
- CI/CD (10%). Branching, committing and pull requests in Databricks Git Folders, and Declarative Automation Bundles (formerly Databricks Asset Bundles) with variables, overrides and the Databricks CLI across dev, test and prod.
- Troubleshooting (10%). Run history, job statuses, Spark UI metrics, Liquid Clustering and predictive optimization, and cluster startup failures, library conflicts and out-of-memory issues.
- Platform (6%). Architecture, Delta Lake, Unity Catalog and compute choices. Short, and a good warm-up before the rest.
Every date we track for this and the other exams we cover is in the exam calendar.