DE Pro Cheat Sheet: Key Facts to Review

The short version

Use this sheet for a final review, not as your only study material. It follows the nine sections of the Databricks exam guide dated October 9, 2026, in our own words. It does not cover everything on the exam, so read the current guide too.

Facts at a glance

Item	Value
Level	Professional
Questions	About 60 scored (guide: 60; certification page: 59)
Time	120 minutes
Format	Multiple choice, no test aids
Passing score	Not published
Validity	2 years
Retake wait	14 days

Section	Weighting
Developing Code (Python and SQL)	23%
Cost & Performance Optimization	15%
Data Ingestion & Acquisition	12%
Data Manipulation	12%
Monitoring and Alerting	10%
Debugging and Deploying	10%
Security and Compliance	8%
Data Governance	5%
Data Modeling	5%

Easily confused pairs

Streaming table vs materialized view. A streaming table processes new data incrementally and suits append-style flows. A materialized view recomputes or maintains a result and reflects updates to the source.
Auto Loader vs COPY INTO. Auto Loader tracks new files incrementally and scales to many files. COPY INTO is a simpler batch-style load.
Sharing vs federation. Sharing publishes your data to recipients. Federation lets you query an external database from Databricks.
Row filter vs column mask. A row filter decides which rows a user sees. A column mask changes what a user sees inside a column.
Deletion vectors vs liquid clustering. Deletion vectors speed up frequent small updates and deletes. Liquid clustering improves selective reads.
Metric view vs materialized view. A metric view holds one governed definition of a metric. A materialized view holds a precomputed result.
Repair run vs full rerun. A repair reruns failed tasks and the tasks that depend on them. A full rerun starts over.
Hashing vs tokenization. Hashing is one way. Tokenization swaps a value for a token that can be mapped back under control.
Drop, fail, and quarantine expectations. Drop discards bad records, fail stops the update, and quarantine keeps the bad records somewhere you can inspect them.

Key concepts by section

Developing Code (23%)

Declarative Automation Bundles (formerly Databricks Asset Bundles) package code and job definitions.
AUTO CDC (formerly APPLY CHANGES) handles change data, including SCD Type 1 and 2.
Stateful streaming needs a checkpoint, and watermarks limit how long state is kept.
Jobs support If/Else and For Each tasks.

Ingestion (12%)

Know Kafka, Kinesis, and Pub/Sub sources, plus cloud storage formats.
Lakeflow Connect handles CDC from databases such as SQL Server, MySQL, and PostgreSQL.

Manipulation (12%)

Window functions, VARIANT with `parse_json` and `variant_get`, AI functions such as `ai_query`.

Monitoring (10%)

System tables and pipeline event logs are queryable tables.

Cost and performance (15%)

Predictive Optimization and managed tables reduce maintenance. The query profile shows skipping, joins, and shuffles.

Security, governance, modeling (18% together)

Least-privilege grants, governed tags for ABAC, and purge for erasure rules.
Privileges granted at a higher level are inherited by objects created later.
Partitioning follows data grain, and clustering follows access patterns.

Decision rules

If the scenario needs full history including late updates, think materialized view.
If bad records must never be lost, think quarantine.
If a failed job should not redo finished work, think repair run.
If a table is slow because of frequent small deletes, think deletion vectors.
If a table is slow on selective reads, think clustering.
If a stem asks for something repeatable or joinable, think system tables or event logs, not a screenshot.
If data must be erased, think about old table versions too.

Common traps

Picking the familiar tool instead of the one the constraint names.
Fixing the wrong symptom.
Trusting study notes from an older outline.
Spending too long on one question. About two minutes each is the average.

Last-day checklist

Re-read the current exam guide.
Test your online proctoring setup on a personal machine, or plan your route to the test center.
Review your weakest two sections only.
Remember that unanswered questions earn nothing, so answer everything.
Sleep.

More guides for DB-DE-PRO

Next step

Go deeper in domains explained or test yourself with sample questions. The exam overview links everything. For more practice, see our practice sets.

Try these practice exams