CareerGuide 14 of 36
Career guide · Guide 14 of 36
Databricks Data Engineer Associate: Exam Guide and Study Plan
Databricks Certified Data Engineer Associate exam guide: format, syllabus and the 7 domain weights, what to study, a 5-week plan and practice questions.
On this page
The Databricks Certified Data Engineer Associate checks that you can do everyday data engineering work on Databricks: load data, transform it with SQL and PySpark, schedule it with Lakeflow Jobs, deploy it with CI/CD, fix it when it breaks, and secure it with Unity Catalog. It is one of the most searched data engineering certifications, and interest in it is growing.
This guide explains the exam in plain words, what each part covers, which free lessons on this site help, and a realistic plan to pass.
The exam at a glance
These facts come from the official Databricks exam page, checked on 31 October 2026.
| Scored questions | 45, multiple choice |
| Time limit | 90 minutes |
| Price | $200 |
| Where | Online (proctored) or at a test centre |
| Aids allowed | None |
| Languages | English, Japanese, Portuguese (Brazil), Korean |
| Prerequisites | None (training is recommended) |
| Valid for | 2 years, then you retake the current version |
| Code in questions | SQL where possible, otherwise Python |
The exam may also include a few unscored questions that Databricks is testing for future use. You cannot tell which ones they are, and extra time is included for them.
What the exam covers
The exam has seven domains. The percentage is the share of questions from each one.
| Domain | Weight | In plain words |
|---|---|---|
| Databricks Intelligence Platform | 6% | What the platform is, the workspace, compute types, and where data and code live. |
| Data Ingestion and Loading | 21% | Getting data in: Auto Loader, COPY INTO, Lakeflow Connect, reading files and tables. |
| Data Transformation and Modeling | 22% | Cleaning and shaping data with SQL and PySpark, Delta tables, MERGE, medallion layers. |
| Working with Lakeflow Jobs | 16% | Scheduling and running work: tasks, dependencies, retries, parameters, repair runs. |
| Implementing CI/CD | 10% | Git folders and Declarative Automation Bundles to move code from dev to prod. |
| Troubleshooting, Monitoring, and Optimization | 10% | Reading the Spark UI and query profile, finding slow steps, fixing small files and skew. |
| Governance and Security | 15% | Unity Catalog: the three-level namespace, grants, managed vs external tables, lineage, sharing. |
Ingestion and transformation together make up 43% of the exam, so spend most of your time there.
What to study for each domain
Platform (6%). Know the control plane and compute plane, the difference between all-purpose, job and serverless compute, and what SQL warehouses are for. Read Databricks for data engineers and workspace, jobs and the lakehouse.
Ingestion (21%). Understand when to use Auto Loader and when COPY INTO (both skip files already loaded), how Auto Loader tracks files and handles new columns, and what Lakeflow Connect does. Read Auto Loader and declarative pipelines and practise Auto Loader vs COPY INTO.
Transformation and modelling (22%). Be fluent in Spark SQL and PySpark DataFrames: joins, aggregations, window functions, handling nulls and nested JSON. Know Delta Lake well: MERGE, schema enforcement and evolution, time travel, OPTIMIZE, and liquid clustering. Understand bronze, silver and gold. Read the PySpark course, the Delta Lake course and medallion architecture.
Lakeflow Jobs (16%). Know how a job is built from tasks, how dependencies and conditions work, how to pass parameters and task values, and what a repair run does. The jobs part of workspace, jobs and the lakehouse covers this.
CI/CD (10%). Know Git folders for version control, and how bundles describe jobs and pipelines in YAML so the same project deploys to dev, staging and prod with databricks bundle deploy.
Troubleshooting and optimisation (10%). Learn to spot skew, spill and too many small files in the Spark UI, and the usual fixes. Read partitions, shuffles and skew and OPTIMIZE, Z-ordering and liquid clustering.
Governance (15%). Know catalog.schema.table names, the grants needed to read a table (USE CATALOG, USE SCHEMA, SELECT), managed vs external tables, volumes, row filters and column masks, lineage and Delta Sharing. Read Unity Catalog governance.
For quick revision, the Top 50 Databricks interview questions cover most of these topics with short answers.
A 5-week study plan
This plan assumes about 1 hour a day and some SQL and Python already. Use the free Databricks trial or Free Edition so you can try everything yourself; hands-on practice is what makes the questions easy.
| Week | Focus | Do this |
|---|---|---|
| 1 | Platform and Delta Lake | Create a workspace, run notebooks, create Delta tables, try MERGE, time travel and OPTIMIZE. |
| 2 | Ingestion | Load files with Auto Loader and COPY INTO; add a new column to the source and watch what happens. |
| 3 | Transformation | Build a small bronze → silver → gold flow in SQL and PySpark; practise joins, windows and JSON. |
| 4 | Jobs, CI/CD and governance | Turn the flow into a multi-task job; break a task and use repair run; create grants in Unity Catalog. |
| 5 | Troubleshooting and review | Read the Spark UI on a slow query, then do the official practice material and revise weak domains. |
In the last week, read the official exam guide again line by line and check you can explain every point in your own words.
Practice questions
These are our own questions, written to show the style and the reasoning. They are not taken from the real exam.
Question 1
New CSV files land in cloud storage all day, millions per month, and a few new columns appear over time. What is the best way to load them into a Delta table?
Show answer
Answer: B. Auto Loader picks up only new files, remembers what it has loaded, scales to millions of files, and can handle new columns. Re-reading the whole folder every day (A) gets slower and creates duplicates.
Question 2
An analyst group needs to read prod.sales.orders. Which privileges do they need?
Show answer
Answer: B. In Unity Catalog you need permission to use each level above the table, plus SELECT on the table itself. C and D give far more access than needed.
Question 3
A MERGE into a Delta table fails because several source rows match the same target row. What is the right fix?
Show answer
Answer: B. MERGE cannot decide which source row should win, so reduce the source to one row per key first, usually the latest one.
Question 4
Task 3 of a 5-task job failed after tasks 1 and 2 succeeded. You fixed the bug. What should you run?
Show answer
Answer: B. A repair run re-runs only the failed task and the tasks that depend on it, so you do not repeat work that already succeeded.
Exam-day tips
- Read the last line of the question first. Many questions describe a scenario, then ask something narrow. Knowing the ask helps you read the scenario well.
- Look for the managed, built-in option. When two answers both work, the exam usually prefers the native Databricks feature (Auto Loader, Unity Catalog, Lakeflow Jobs) over custom code.
- Leave time to review. Answer the questions you are sure of first, then come back to the flagged ones.
- Flag and move on. 45 questions in 90 minutes is about 2 minutes each. Do not get stuck.
A word on “dumps”
Some websites sell lists of “real exam questions”. Using them breaks the Databricks exam agreement and can get your certificate revoked, and the questions are often wrong or out of date. The certificate is only worth something if the skills are real, so learn the platform hands-on and use the official practice material.
Is it worth it?
The certificate helps most when you are moving into data engineering, or into a team that uses Databricks, because it shows you know the platform’s current features. It does not replace experience: interviewers will still ask you to explain a pipeline you built. The best combination is the certificate plus one small project you can talk about in detail.