DP-750 · Microsoft
Studying for DP-750? Two changes you need to know about
DP-750 changes on 19 October, and two of the changes touch what you study. What they are, what the four skill areas weigh, and the objectives where you must choose the right tool.
Microsoft DP-750 covers data engineering on Azure Databricks: integrating and modelling data, building and deploying pipelines, and keeping workloads healthy, with governance in Unity Catalog running through all of it.
The study guide expects you to ingest and transform data with SQL and Python, to work with software development lifecycle practices including Git, and to be familiar with Microsoft Entra, Azure Data Factory and Azure Monitor.
What the exam measures
| Skill area | Weight |
|---|---|
| Set up and configure an Azure Databricks environment | 15 to 20% |
| Secure and govern Unity Catalog objects | 15 to 20% |
| Prepare and process data | 30 to 35% |
| Deploy and maintain data pipelines and workloads | 30 to 35% |
Two areas carry most of the exam: processing data and running the pipelines that do it. The two lighter areas are where Unity Catalog lives, and it reaches into the heavy ones too.
Which version you will sit
- 11 March 2026: the version of the objectives the English exam measures until 19 October.
- 19 October 2026: the date Microsoft updates the English exam to a new version, already published in the study guide.
Unlike the Fabric exams, DP-750 has two changes worth knowing in October:
- Deletion vectors leave the clustering objective. It will name only liquid clustering and Z-ordering.
- Databricks Asset Bundles become Declarative Automation Bundles, and the deployment objective refers to the "Databricks CLI". The skill (packaging and deploying a bundle) is the same.
The weights do not move. The full comparison is in what actually moves on 19 October.
How the exam is scored and delivered
- Time: 120 minutes.
- Scoring: Microsoft requires a score of 700 or greater.
- Language: English. If English is not your preferred language, you can request an extra 30 minutes.
- Delivery: proctored, and it may include interactive components.
- Retakes: a second attempt can be taken 24 hours after the first; the wait before later attempts varies.
The objectives where a choice is the question
- Compute. Job compute, serverless, warehouse, classic compute or shared compute, then the settings that follow: node type, cluster size, autoscaling, pooling, Photon acceleration and runtime version.
- Ingestion tool and method. The tool: Lakeflow Connect, notebooks or Azure Data Factory. The SQL method: CREATE TABLE AS (CTAS), CREATE OR REPLACE TABLE or COPY INTO. Then change data capture feeds, Spark Structured Streaming, Azure Event Hubs, and Lakeflow Spark Declarative Pipelines with Auto Loader.
- Table design. Managed or unmanaged tables, a partitioning scheme, a clustering strategy, the type of slowly changing dimension, and a temporal (history) table to record changes over time.
- Troubleshooting. Caching, skewing, spilling and shuffle issues, investigated with the DAG, the Spark UI and the query profile. Know which tool shows which problem.
Where else people slip
- Security has many layers. Table- and column-level access control and row-level security, attribute-based access control with tags and policies, service principals and managed identities, Azure Key Vault secrets, lineage, audit logging and a secure strategy for Delta Sharing.
- Data quality is explicit. Validation checks for nullability, data cardinality and range, data type checks, schema enforcement and schema drift, and pipeline expectations.
- The lifecycle is examined, not assumed. Git branching and pull requests, a testing strategy that runs from unit tests to user acceptance testing, and bundles deployed with the CLI and with REST APIs.
A study plan that follows the guide
- Prepare and process data (30 to 35%). Ingestion tools and methods, profiling, transformations (joins, unions, pivots and the like), the data quality list above, and loading with merge, insert or append.
- Deploy and maintain pipelines and workloads (30 to 35%). Lakeflow Jobs (triggers, schedules, alerts, automatic restarts), the development lifecycle, troubleshooting, OPTIMIZE and VACUUM, and log streaming with Log Analytics in Azure Monitor.
- Set up the environment (15 to 20%). Compute choices and settings, then catalogs, schemas, volumes, tables, views and a foreign catalog through connections.
- Secure and govern Unity Catalog (15 to 20%). The security and governance list above.
If you sit the exam after 19 October, learn the bundle material under its new name and treat deletion vectors as background rather than as a named objective.
Every date we track for this and the other exams we cover is in the exam calendar.