Free study guide · Databricks
How to pass the Databricks Certified Data Engineer Associate
What the exam covers, how to plan your study, and how to read its questions. From the CertLumen course for this exam, kept up to date with the official exam guide.
- Questions
- 45
- Time
- 90 minutes
- Passing score
- Not published by Databricks; this kit targets ≥ 80% on unseen mocks
- Exam fee
- USD 200 (plus applicable taxes)
- Valid for
- 2 years (renew by taking the current exam or the 25-question renewal exam)
- Delivery
- Online proctored or test center
What the exam covers
A study plan that works
The default path for engineers with some Spark or SQL experience and up to about six months on Databricks. It covers all 41 lessons and 12 labs in order, with a domain drill after each module and daily flashcards. Mock Exam 1 on day 20 is a diagnostic, Mock Exams 2–3 measure progress, and Mock Exam 4 plus weak-area review closes it out. Day 30 is a light day, so you sit the exam on day 31. The heaviest sections (d3 22%, d2 21%, d4 16%, d7 15%) get the most days. Plan on about 1.5–2 hours a day.
| Day | Focus | Time |
|---|---|---|
| 1 | Kick-off: exam map, lakehouse and Spark basics | 100 min |
| 2 | DataFrames and Spark SQL | 110 min |
| 3 | Delta Lake and Structured Streaming | 120 min |
| 4 | Platform architecture and Unity Catalog | 90 min |
| 5 | Compute, cost models and the workspace | 110 min |
| 6 | Ingestion patterns and COPY INTO | 100 min |
| 7 | Auto Loader deep dive | 110 min |
| 8 | Lakeflow Connect, JDBC and REST | 110 min |
| 9 | Semi-structured data and choosing a method | 120 min |
| 10 | Ingestion review day | 100 min |
| 11 | Medallion cleaning and joins | 110 min |
| 12 | Columns, deduplication and aggregations | 120 min |
| 13 | Tuning parameters and Gold objects | 110 min |
| 14 | Declarative pipelines and data quality | 120 min |
| 15 | Transformation review day | 100 min |
| 16 | Lakeflow Jobs fundamentals and control flow | 110 min |
| 17 | Triggers and orchestration choices | 120 min |
| 18 | Jobs drill and Git folders | 110 min |
| 19 | Bundles and the Databricks CLI | 120 min |
| 20 | Mock Exam 1 (diagnostic) | 120 min |
| 21 | Monitoring jobs and the Spark UI | 110 min |
| 22 | Liquid clustering, predictive optimization and failures | 120 min |
| 23 | Managed vs external tables and privileges | 120 min |
| 24 | Row filters, column masks and ABAC | 120 min |
| 25 | Lineage, auditing and governance review | 110 min |
| 26 | Mock Exam 2 | 120 min |
| 27 | Weak-area remediation | 110 min |
| 28 | Mock Exam 3 | 120 min |
| 29 | Mock Exam 4 and final weak-area review | 120 min |
| 30 | Light day: cheat sheet and rest | 60 min |
Exam strategy
How the exam works
The Databricks Certified Data Engineer Associate exam checks whether you can use the Databricks Data Intelligence Platform for foundational data engineering: ingesting, transforming, orchestrating, deploying, troubleshooting and governing data. Every item is single-answer multiple choice, mostly short scenarios asking for the best option under a stated constraint. Knowing that a feature exists isn't enough; you have to know when it's the right choice.
| Fact | Detail (verified on the exam page, exam guide and certification FAQ) |
|---|---|
| Scored questions | 45 multiple-choice questions |
| Unscored items | May be included to gather statistics. They aren't marked and don't affect your score, and extra time is built in for them. |
| Time limit | 90 minutes, about 2 minutes per question |
| Fee | USD 200 plus applicable taxes |
| Delivery | Online proctored or at a test center. You register on Webassessor, and Kryterion delivers the exam. |
| Test aids | None allowed: no documentation, notes, or second screen |
| Languages | English, Japanese, Portuguese (BR) and Korean, per the exam page |
| Prerequisites | None required. Training and about six months of hands-on Databricks experience are highly recommended. |
| Passing score | Not published. Databricks says it can change. |
| Validity | 2 years. Renew by passing the current live exam, or the 25-question renewal exam, which opens 2 months before your credential expires. |
You can't tell which items are unscored, so treat every question as if it counts. If a question seems to be outside the outline, answer it quickly with your best guess and move on. It may be one of the unscored items.
Section weights: where the points are
The weights tell you where to spend study time. Four sections carry 74% of the score: Ingestion, Transformation, Lakeflow Jobs and Governance. The question counts below are estimates from the published weights. Databricks doesn't publish per-section counts, and individual exam forms vary.
| Section | Weight | ≈ Questions | Kit module | Highest-yield topics |
|---|---|---|---|---|
| 1. Databricks Intelligence Platform | 6% | 3 | d1 (+ d0) | Delta Lake + Unity Catalog as the foundation, compute selection, serverless limits, cost models |
| 2. Data Ingestion and Loading | 21% | 9 | d2 | Auto Loader options and schema evolution, COPY INTO idempotency, Lakeflow Connect managed vs standard, method selection, JSON/VARIANT |
| 3. Data Transformation and Modeling | 22% | 10 | d3 (+ d0) | Joins and unions, dedup, aggregations, tuning configs, Gold objects, declarative pipelines, expectations |
| 4. Working with Lakeflow Jobs | 16% | 7 | d4 | Run if conditions, If/else and For each tasks, retries, task values, scheduled/file arrival/table update triggers |
| 5. Implementing CI/CD | 10% | 4–5 | d5 | Git folder workflow, bundle variables, targets and modes, databricks bundle CLI |
| 6. Troubleshooting, Monitoring, and Optimization | 10% | 4–5 | d6 | Run history baselines, Spark UI skew/shuffle/spill, liquid clustering, predictive optimization, OOM and library conflicts |
| 7. Governance and Security | 15% | 7 | d7 | Managed vs external tables, GRANT/REVOKE and the privilege hierarchy, row filters, column masks, ABAC |
| Total | 100% | 45 |
Spark veterans usually do well on Section 3 and lose points on Sections 2, 4 and 5, which test newer products: Lakeflow Connect managed connectors, table update triggers, For each tasks, Run if conditions, bundle variables and targets. Budget study time by weight and by unfamiliarity.
How to use this kit
- Learn. Read the lesson, answer its self-checks, and write down any key point you would have missed.
- Lab. Do the lab on Databricks Free Edition. Some exam details only stick once you've seen them happen, like Auto Loader failing once on a new column or COPY INTO loading 0 rows on a re-run.
- Drill. After each module, take its domain drill untimed. Read every explanation, including the notes on the options you rejected. The distractors teach as much as the answer.
- Flashcards. Review in short daily sessions. Mark the cards you miss and repeat them until you get them right two days in a row.
- Mocks. Take the four mock exams (45 questions, 90 minutes each) under exam conditions. Never take them open-book.
- Review loop. Keep an error log with four columns: topic, why you missed it (knowledge gap, misread constraint, or trap), the correct rule, and where the rule lives in the kit. Re-read that section and retest similar questions 48 hours later.
Your score on a repeated mock measures memory, not readiness. Count only first attempts toward the readiness criteria, and use repeat attempts just to review explanations.
Readiness criteria
- ≥ 80% on two mock exams you hadn't seen before, each taken timed in one sitting.
- ≥ 70% in every domain on those mocks. One weak domain carrying 15–22% of the weight can sink an otherwise good attempt.
- You finish a mock in ≤ 75 minutes and still have time for a review pass.
- Without notes, you can rebuild these cheat-sheet tables: Auto Loader vs COPY INTO, Run if conditions, Gold objects, UC privileges, and Spark UI symptom → fix.
- You've done every lab at least once and can explain what each lab proved.
Databricks doesn't publish the passing score, and community reports usually put it around 70%. Aiming for 80% on unseen mocks leaves a margin for a harder exam form, unfamiliar wording and exam-day nerves.
Time management
You have 90 minutes for 45 scored items plus any unscored ones, or roughly 2 minutes per question. Recall questions take 30–60 seconds, which saves time for long scenario and code questions. Use three passes.
- First pass (≈ 55–60 min). Answer everything you can within about 90 seconds. If you're stuck between two options, pick the better one, flag the question, and move on. Never leave a question blank.
- Second pass (≈ 15–20 min). Go back to the flagged questions. Re-read the stem's constraint words before re-reading the options.
- Final check (≈ 5 min). Make sure nothing is unanswered. Change an answer only for a concrete reason, such as a constraint you missed, and never on a vague feeling.
| Checkpoint | Target elapsed time |
|---|---|
| Question 15 | ≤ 25 min |
| Question 30 | ≤ 50 min |
| Question 45 | ≤ 70 min, then flagged questions only |
An unanswered question earns nothing. When time is short, answer every remaining question first, then go back to the flagged ones.
Reading scenario questions
A typical item gives a few sentences of context, some of it irrelevant, and then asks for the best solution. One or two constraint words usually decide the answer. Read the question in this order:
- Read the last sentence (the actual ask) first.
- Find the constraint words: cost, overhead, latency, governance, privileges, idempotency.
- Predict the answer before you look at the options.
- Eliminate every option that breaks any constraint, then pick from what's left.
| Constraint wording | What it usually rewards |
|---|---|
| "least operational overhead", "minimal maintenance" | Managed, serverless, declarative options: Lakeflow Connect managed connectors, Lakeflow Spark Declarative Pipelines, serverless compute, predictive optimization, CLUSTER BY AUTO |
| "most cost-effective" | Jobs compute or serverless jobs instead of all-purpose clusters, availableNow batches instead of always-on streams, autoscaling with auto-termination, spot workers for fault-tolerant batch |
| "incrementally", "only new files" | Auto Loader or COPY INTO, which both track processed files, and streaming tables |
| "near real-time", "low latency" | Structured Streaming with a processing-time or continuous trigger, continuous pipelines or jobs. Not an hourly schedule. |
| "governed", "in Unity Catalog" | UC managed tables, volumes instead of DBFS, external locations with storage credentials, connectors that land data in UC |
| "minimum privileges", "least privilege" | The smallest set of grants at the lowest level that works, such as USE CATALOG + USE SCHEMA + SELECT. Never ALL PRIVILEGES or ownership. |
| "idempotent", "safe to re-run" | COPY INTO, Auto Loader checkpoints, MERGE instead of INSERT |
| "schema changes frequently" | Auto Loader schema evolution (addNewColumns or rescue), mergeSchema, the VARIANT type |
| "same code across dev/test/prod" | Declarative Automation Bundles with variables and target overrides |
| "as soon as data arrives" | Data-driven triggers (file arrival or table update) instead of cron schedules |
Elimination tactics
Words that point to the answer
Phrases in a question that usually point to one answer. The course has the full list.
| When the question says | Think of |
|---|---|
| reliable rollback after a bad ingest, audit trail, single source of truth for AI and BI | Delta Lake ACID transactions and time travel, governed by Unity Catalog |
| many analysts running ad hoc SQL concurrently, fast startup, cost-aware | Serverless SQL warehouse (autoscaling) |
| scheduled production ETL at the lowest cost | Serverless jobs compute or jobs compute (job cluster), not all-purpose |
| workload needs the RDD API, R, init scripts or arbitrary Spark configs | Classic compute in dedicated access mode |
| several users share one cluster with Unity Catalog fine-grained access control | Standard access mode (formerly shared) |
| query a PostgreSQL or Snowflake database without copying the data | Lakehouse Federation (connection + foreign catalog) |
| share live tables with a partner organization, recipient may not use Databricks | Delta Sharing (open sharing protocol) |
| who read or changed a table last week | Audit logs: the system.access.audit system table |
| which downstream dashboards break if this column changes | Unity Catalog data lineage |
| incrementally load new files from cloud storage with schema evolution | Auto Loader (cloudFiles) |
Revise from the cheat sheet
The course's cheat sheet has a page for each of these:
- Delta Lake commands and defaults
- Structured Streaming essentials
- Compute selection matrix
- Auto Loader options
- COPY INTO options
- Ingestion decision matrix
- Spark DataFrame API ↔ SQL equivalents
- Joins and set operations
- Aggregations and deduplication
- Tuning configs and defaults
- Gold layer objects compared
- Declarative pipeline syntax and expectations
- Lakeflow Jobs: tasks, Run if, triggers
- Bundle databricks.yml skeleton and CLI commands
- Git folder operations
- Spark UI: symptom → cause → fix
- Liquid clustering and predictive optimization
- Unity Catalog privileges and hierarchy
- Row filters, column masks and ABAC
- Managed vs external tables and volumes
Practise with the real exam's mix
395 exam-style questions, 4 timed practice exams, flashcards and an AI tutor.