Databricks Certified Data Engineer Associate (DEA) — exam overview¶

This page is the map. It tells you what the exam is made of, what its questions look like, how to spend the 90 minutes, and where the material you find elsewhere will mislead you.

That last part matters more on this exam than on most. The current version is a substantial rewrite, and almost every course, blog post and practice set still describes the previous one.

The exam at a glance¶

Fact	Value
Scored questions	45, all multiple choice
Unscored questions	Up to 10, not identified, so you may see up to 55 in total
Time limit	90 minutes, with the time for unscored items already included
Registration fee	USD 200, plus local tax
Delivery	Online proctored, or at a test centre
Test aids	None permitted
Prerequisite	None, though six months of hands-on Databricks is recommended
Validity	Two years, then the Data Engineer Associate Renewal Exam or the current full exam
Languages	English, Japanese, Brazilian Portuguese, Korean

The unscored items look exactly like the scored ones, and the guide says up to 10 may be included. So 45 questions decide your result, but you should plan your time for 55. That is why our full-length practice tests have 55 questions in 90 minutes.

The passing score is not published, and you will be told otherwise¶

Databricks does not publish passing scores for any certification. Their own certification FAQ says the scores are set by statistical analysis, change as exam forms are updated, and are therefore not published.

Two corrections follow, and both are worth carrying:

70% is not an official figure. It is repeated across third-party sites and it has no published source behind it.
80% is a real Databricks number for a different thing. It is the minimum for Badges, which are free, unproctored assessments. It is not the certification pass mark.

The scoring model is likewise unstated. Databricks does not say whether a per-domain minimum applies, so do not plan to write off a domain and make up the marks elsewhere. Prepare as though every section counts, because nothing published says it does not.

🎯 Ready to test yourself? The DataCertLab Databricks Data Engineer Associate Practice Tests on Udemy are built from the same exam blueprint as these notes, with a full explanation for every question.

🎟️ Free for the first 100 learners: use code DATABRICKS-DEA-100 until 1 December 2026. Apply it on Udemy

What the exam covers¶

Seven sections, and their weights are published for the first time in this version. Study time follows this table.

Section	Weight	What it actually rewards
Databricks Intelligence Platform	6%	Naming the platform's parts, and picking the right compute for a workload
Data Ingestion and Loading	21%	Choosing between ingestion paths, and saying why one beats the others
Data Transformation and Modeling	22%	PySpark and SQL transformation, and which Gold-layer object to build
Working with Lakeflow Jobs	16%	Orchestration, task dependencies, and trigger choice
Implementing CI/CD	10%	Bundles, Git folders, and the command-line workflow
Troubleshooting, Monitoring, and Optimization	10%	Reading the Spark UI and run history, and diagnosing from evidence
Governance and Security	15%	Unity Catalog privileges, masking, row filters, and attribute-based policies

Two sections carry nearly half the exam between them: Data Ingestion and Loading at 21%, and Data Transformation and Modeling at 22%. If your time is short, that is where it goes.

Section names above are spelled as Databricks spells them, including American spellings such as Modeling. The rest of this material uses British spelling.

What changed in the May 2026 version¶

The previous exam guide was published in July 2025. It had five sections and no published weights. This one has seven sections, 33 objectives, and weights for every section.

Four features were renamed. The guide leads with the new name and parenthesises the old one — it writes "formerly Databricks Repos" and "formerly Databricks Asset Bundles" in the objectives themselves. So both names are in the current guide, and learning only the old one leaves you reading the parenthesis instead of the objective.

Old name	Current name in the guide
Workflows	Lakeflow Jobs
Delta Live Tables (DLT), LDP	Lakeflow Declarative Pipelines
Databricks Asset Bundles (DABs)	Declarative Automation Bundles
Databricks Repos	Databricks Git folders

The guide was refreshed on 6 August 2026, and the exam was not. The update table on its first page says the exam was published on 4 May 2026 and the guide updated on 6 August. The sections, objectives and weights are the same. What moved:

Names tightened: "Databricks Platform" instead of "Databricks Data Intelligence Platform", "Lakeflow Declarative Pipelines" instead of "Lakeflow Spark Declarative Pipelines", and "Declarative Automation Bundle" instead of "Automation Bundle".
The guide now says up to 10 unscored items may appear, and that you can recertify with a renewal exam.
The sample questions went from five to ten, with an official answer key and explanations. Two of the original five changed their answer, which the next section covers.

Two sections are entirely new. Implementing CI/CD at 10% and Troubleshooting, Monitoring, and Optimization at 10% had no counterpart section in the previous guide. Together they are a fifth of the exam, and older study material does not cover them as sections at all.

New topics appear with no previous counterpart. Lakeflow Connect and its standard and managed connectors, the COPY INTO command, Unity Catalog attribute-based access control policies, column masking and row-level security, Liquid Clustering, predictive optimization, the three trigger types, and the Databricks command-line interface.

And several topics were dropped from the outline entirely. This is the trap, because it is where older courses spend their time:

Delta Sharing, which had four separate objectives in the July 2025 guide
Lakehouse Federation
Databricks Connect
Unity Catalog key roles, audit-log storage, and lineage

None of these has an objective of its own in the current outline.

But "no longer its own objective" is not "off the exam", and the guide proves it. Sample question 3 in the current guide asks how Databricks audit logs are stored — the delivery format, the latency, whether files get overwritten. Audit-log storage was a July 2025 objective and is not one now. The guide files that question under objective 2.1, "Enable and detail data ingestion patterns", which absorbed it.

So the rule to carry in is about the objective, not the topic:

A topic without its own objective can still be the right answer, where a current objective covers the ground. Audit-log storage under an ingestion objective is the guide's own worked example.
Delta Sharing and Lakehouse Federation are not named as objectives anywhere in the current outline. That is all the outline establishes. It does not put them beyond every question: objective 1.1 asks about the platform's core components "such as its architecture, Delta Lake, and Unity Catalog", and a such as list is open at the end.

Ask which objective a question is testing, not whether its topic still has a bullet. Databricks publishes nothing about how it builds wrong answers, so treat none of this as a rule about where distractors come from.

One product has three names in play¶

The guide names Lakeflow Declarative Pipelines (the May version said Lakeflow Spark Declarative Pipelines). Current Databricks documentation calls the same product Lakeflow pipelines — the page on what happened to Delta Live Tables says the product "has been updated to Lakeflow pipelines". The guide's own training list names the open-source framework the product extends, Apache Spark Declarative Pipelines.

This does not change the exam answer. The exam guide is what the exam is written from. If an option says Lakeflow Declarative Pipelines and another says Delta Live Tables, the first is current and the second is the retired name. Do not talk yourself out of a correct option because the documentation you just read used a shorter name. Learn the names as one product.

What the questions actually look like¶

Databricks publishes sample questions inside the exam guide, and there is no official practice exam for this certification, whatever older links suggest. What follows comes from working through every sample Databricks publishes for this exam, read alongside the samples in its other current exam guides.

There are ten of them now, with answers explained. The August 2026 guide prints ten samples, each under the objective it illustrates, followed by an answer key that explains why each wrong option is wrong. They are the best evidence there is of how Databricks writes for this exam. Read the explanations as well as the keys: they show the reasoning the exam expects.

Two answers changed, and older material still teaches the old ones.

The Spark UI skew sample is now a groupBy aggregation with adaptive query execution already on, and the answer is to salt the aggregation key. The explanation says why: AQE's skew handling splits skewed joins, not aggregations.
The sample about analysts running ad hoc SQL all day now answers with a serverless SQL warehouse, where the May version said a high-concurrency cluster.

The published samples are scenarios. Each opens with a situation rather than a question, and the pattern holds across Databricks' other current exams too. Treat that as a consistent house style rather than a published rule: prepare for scenarios, without assuming bare recall is impossible.

Four options, one correct, on this exam's samples. Every DEA sample gives options A to D with a single answer. The format is not guaranteed — another current Databricks associate exam includes a five-option item asking for two answers — but if a stem wants two, it says so. No sample anywhere uses "All of the above" or "None of the above".

Most stems are short, and some are not. The ten samples run from about 24 to 110 words, clustering near 50. The long one is the Spark UI question, and every extra word in it is evidence you need. Where a stem runs long, the extra words are evidence rather than padding.

The decisive constraint is in lower case. No published sample capitalises MOST, LEAST or BEST, and Databricks publishes no rule about it either way — so treat it as an observed habit rather than a guarantee. The constraint sits quietly in the closing question instead — "with the lowest cost", "without rearranging the data", "while emphasizing simplicity" — or as a stated assumption. Read the last line twice. If you are used to exams that shout the qualifier, do not wait for the shout.

Expect code. The certification page says code is given in SQL where possible and in Python otherwise. The newer samples show it in the options: a BUNDLE_VAR_ environment variable, a masking function built on is_account_group_member, ALTER TABLE ... UNSET MANAGED. In Databricks' other current guides it also sits inside the scenario — a CREATE TABLE statement or a few lines of PySpark. Prepare to read code in a stem and in the options.

Numbers in a stem are load-bearing. Where a sample gives a shuffle read size or a task duration, the answer turns on it. The samples do not decorate a stem with a figure that does not matter.

Timing and pacing¶

You get 90 minutes for up to 55 questions: 45 scored and up to 10 unscored that you cannot tell apart. Plan on a little over a minute and a half each, and treat every question as one that counts.

That pace is comfortable, given a median stem of about 50 words. The risk is not running out of time. It is spending four minutes on a governance question you half-know and then rushing three ingestion questions you would have got right.

A workable pass structure:

First pass, about 60 minutes. Answer everything you know. Flag anything that needs more than roughly a minute and a half and move on immediately.
Second pass, about 25 minutes. Return to the flagged questions with the pressure off.
Last five minutes. Confirm nothing is unanswered.

Answer every question, including ones you run out of time on. This is practical advice, not a scoring fact. Databricks publishes nothing about how a blank is treated or whether a wrong answer is penalised — it does publish some scoring facts, such as unscored items not affecting your score, but not these two. Nothing on record makes an attempted answer worse than a blank, so answer.

How to study for this exam¶

The whole route, before the detail. The part most people get wrong is the loop at the bottom: a failed mock sends you back to specific objectives, not back to another mock.

flowchart TD
    A["Read this<br/>overview"] --> B["Study one<br/>section"]
    B --> C["Do its hands-on<br/>exercises"]
    C --> D{"All seven<br/>done?"}
    D -->|"No"| B
    D -->|"Yes"| E["Sit a<br/>full mock"]
    E --> F["List the objective<br/>behind each miss"]
    F --> G{"Scored 32,<br/>nothing shaky?"}
    G -->|"No"| H["Restudy those<br/>objectives"]
    H --> E
    G -->|"Yes"| J["Book the exam"]

Work through the sections in the guide's order, not by weight. The platform section is only 6%, but it defines the vocabulary every later section assumes. Read it first even though it is worth least.

Weight your depth, not your order. Ingestion and transformation together carry almost half the marks. Governance carries 15% and is heavily rules-based, which makes it the cheapest section to convert into reliable marks.

Get hands on. Databricks Free Edition costs nothing, and this exam asks what happens when you do things rather than what the documentation says. The checklist below is the minimum.

Check the names before you trust any source. A source using Workflows, Delta Live Tables, Asset Bundles or Repos is a source to check, not necessarily an old one — a recent tutorial may use a former name deliberately, for readers who learned it that way, and Databricks itself keeps the old names in compatibility notes. Treat the old name as a prompt to confirm the material against the current guide, and learn the name the guide leads with.

Read the objectives as written. The guide's own objective list is the syllabus, and it is specific. Where an objective names a command, learn that command.

The hands-on checklist¶

These all run in a Free Edition workspace. Each answers a family of questions, and the section it serves is in the last column.

Do this	What it settles	Section
Create a managed table and an external table, then drop each	What `DROP` does to the underlying files	Governance and Security
Load the same files with `COPY INTO`, then with Auto Loader	Why one is re-runnable and the other tracks state	Data Ingestion and Loading
Add a column to the source, rerun Auto Loader	Schema evolution, and what happens when it is off	Data Ingestion and Loading
Build a view, a materialized view and a streaming table over one source	Freshness against recompute cost	Data Transformation and Modeling
Chain three tasks in a job, make the middle one fail, then repair the run	Task dependencies, and what repair reruns	Working with Lakeflow Jobs
Set a file-arrival trigger and drop a file into the location	How data-driven triggers differ from schedules	Working with Lakeflow Jobs
Deploy the same bundle to two targets with different variables	Environment overrides and promotion	Implementing CI/CD
Grant a group `SELECT`, then apply a column mask and a row filter	Which control acts where	Governance and Security

One exercise Free Edition cannot give you¶

Objective 6.3 asks you to spot skew, shuffle and spill by reading stage-level metrics in the Spark UI. You cannot practise that in Free Edition. Free Edition runs on serverless compute only, and on serverless the Spark UI is not available — the documentation says to use the query profile instead.

That leaves two honest options:

If you have	Do this	What you get
Free Edition only	Run a deliberately skewed join, then open the query profile	The slowest and most expensive operator, with its time, rows, memory and shuffle activity
A workspace with classic compute	Run the same join on a classic cluster, then open the Spark UI stage detail	The view the objective names: per-task minimum, median and maximum shuffle read

Be clear about the difference, because it is the difference the objective turns on. The query profile works at the level of operators — it will show you that one join is consuming the time and moving the data, which is worth seeing. It does not break a stage down into its individual tasks, so it cannot show you the lopsided spread across tasks that identifies skew. That comparison lives in the Spark UI. If you can borrow a workspace with classic compute for an hour, spend it here.

Official resources¶

This list is deliberately short, and all of it is first-party. Most of it is free; the cost column says where that stops.

Resource	Why it is worth your time	Cost
Exam guide PDF	The syllabus, ten sample questions with explained answers, and the only authority on what is asked	Free
Certification page	Registration, languages, and the link to the current guide	Free
AI Prep Guide	Databricks' own method for studying with an AI tool, including how to catch the outdated answers they give	Free
Databricks Academy	The courses the guide's Recommended Training names	Self-paced free; instructor-led is paid, and access can depend on entitlement
Product documentation	The product truth, cloud-tabbed — check you are on your cloud's tab	Free
Free Edition	Where the hands-on checklist gets done, with the one gap noted above	Free
Certification FAQ	Scoring policy, retakes, and what is not published	Free

Check the certification page for a newer guide about two weeks before you sit. Databricks updates it whenever the exam changes, and this exam changed substantially in May 2026.

Are you ready? A readiness measure, not a prediction¶

Read this first, because it decides how to use everything below. Databricks does not publish a pass mark, so no mock score predicts your result. Databricks says so itself, in its own prep guide: practice questions "aren't a calibrated readiness check; your real signal is full coverage of the exam guide objectives plus hands-on confidence."

Take that seriously, because it changes what a mock is for. A mock is a diagnostic, not a verdict. Use it to find which objectives you cannot answer, then go and learn those. A score that tells you "probably ready" has told you nothing you can act on.

So there is one number here, and only one.

The threshold: 39 correct out of 55 on a full-length mock — 55 questions, 90 minutes, no notes. That is this course's own readiness bar, the same figure shown on our practice tests. It is not a Databricks pass mark, and it is not the 70% myth corrected earlier on this page; it is a threshold we chose, and 39 of 55 is what it works out to. Below it, keep studying. Above it, you are in reasonable shape — and still not finished, because the real signal is objective coverage.

Then read the mock by section, because the total hides where the marks went. A 55-question mock splits by the published weights, so each section contributes a fixed number of questions:

Section	Questions in a 55-question mock
Databricks Intelligence Platform	3
Data Ingestion and Loading	12
Data Transformation and Modeling	12
Working with Lakeflow Jobs	9
Implementing CI/CD	6
Troubleshooting, Monitoring, and Optimization	5
Governance and Security	8

Two things fall out of that table. Dropping half of Data Transformation and Modeling costs you six marks, while getting every Platform question wrong costs three — so a weak section is only worth worrying about in proportion to its size. And because Databricks does not say whether a per-section minimum applies, do not let any section sit near zero on the assumption you can carry it elsewhere.

Then convert the section into objectives, which is the step that actually works. Take each question you got wrong, find the objective it tests in the guide's outline, and mark that objective. Objectives with two or more misses are your study list. Full coverage of that list, plus the hands-on checklist above, is the signal Databricks itself points at — and unlike a score, it is something you can finish.

Retake the same mock only to check a fix. A score that climbs because you remember the items is measuring memory, not readiness.

The last 24 hours¶

Stop learning new material. Reread the section weights and confirm you know which two sections carry the most marks. Review the renamed features until the current name comes first. Read the list of topics that lost their own objective once, so that meeting one in the exam prompts you to ask which current objective the question is testing. Then rest, because this is a 90 minute exam and reading accuracy is what it measures.

This page tracks the official exam guide version August 6, 2026, for the exam published 4 May 2026. Databricks updates the guide whenever the exam changes. Check the certification page for a newer version about two weeks before you sit.

🎯 Practice for the Databricks Data Engineer Associate exam¶

The DataCertLab Databricks Data Engineer Associate Practice Tests are built from the same blueprint as these notes, and every question comes with a full explanation. Read a domain, test yourself on it, then go back to whatever your wrong answers point at.

👉 Take the practice tests on Udemy

🎟️ Free for the first 100 learners: use code DATABRICKS-DEA-100 until 1 December 2026. Apply it on Udemy

💡 Tip: Open the link in a new tab to keep these notes open for reference.

Learn → Practice → Review → Improve

All certifications