Below are twelve scenario-based practice questions for DP-600 (Implementing Analytics Solutions Using Microsoft Fabric), the exam behind Microsoft Certified: Fabric Analytics Engineer Associate, split into Easy, Medium, and Hard tiers — four questions each — with a full rationale for the correct choice and a specific reason every distractor is wrong. Domain names referenced below come from Microsoft's official DP-600 study guide, checked September 7, 2026.
How to Use These Difficulty Tiers
Start with Easy. If you miss more than one, revisit that question's domain in the DP-600 study guide before moving on. Medium adds a single complicating factor to each scenario; Hard requires ruling out two or three plausible-looking causes the way a real case-study question does. Getting most of Hard right on a first pass is a strong signal you are close to exam-ready.
Easy
Questions 1–4Recognize the right Fabric feature or concept for a directly stated need.
Prepare Data
Question 1: A team needs to ingest continuous IoT sensor readings and query them with low latency as they arrive. Which Fabric item is the best fit?
- A. A lakehouse, since it can store any file format
- B. An Eventhouse, built for real-time and streaming data
- C. A warehouse, since it supports T-SQL
- D. A Power BI Desktop project (.pbip)
Why B is correct: An Eventhouse is Fabric's purpose-built store for real-time and streaming data, queried with KQL — exactly the low-latency, continuously-arriving data pattern described.
Why A is wrong: A lakehouse can store many file formats but is not optimized for low-latency streaming ingestion and query the way an Eventhouse is.
Why C is wrong: A warehouse is built for structured, T-SQL-queried analytical data at rest, not continuously arriving streaming events.
Why D is wrong: A .pbip project is a Power BI Desktop source-control artifact, not a data store — it has no role in ingesting streaming data.
Implement and Manage Semantic Models
Question 2: A semantic model author wants viewers to always see the latest lakehouse data without manually refreshing the model. Which storage mode fits without any additional configuration?
- A. Import, since it is the fastest storage mode
- B. DirectQuery, because it is the newest storage mode
- C. Direct Lake, which reads Delta tables directly without an import/refresh cycle
- D. Any mode works identically for this requirement
Why C is correct: Direct Lake reads OneLake Delta tables directly, so report viewers see current data without a scheduled import refresh — the specific behavior the scenario asks for.
Why A is wrong: Import requires a scheduled or manual refresh to reflect new data — it does not update automatically as source data changes.
Why B is wrong: DirectQuery does query live data, but it is not the mode described as reading OneLake Delta tables directly without an import; being "newest" is also not a real selection criterion.
Why D is wrong: The storage modes behave differently on freshness and performance, so the choice is not interchangeable here.
Maintain a Data Analytics Solution
Question 3: A workspace admin wants to prevent a Contributor from opening one specific warehouse in an otherwise shared workspace. What should they configure?
- A. Change the user's workspace role to Viewer for everyone
- B. A sensitivity label on the workspace
- C. Row-level security on the warehouse's tables
- D. Item-level access control on that specific warehouse
Why D is correct: Item-level access control lets an admin restrict access to one specific item without changing the user's broader workspace role or affecting other items.
Why A is wrong: Downgrading the workspace role affects everything the user can do in the workspace, not just the one warehouse — broader than the requirement.
Why B is wrong: A sensitivity label classifies content for compliance purposes; it does not enforce access control on its own.
Why C is wrong: Row-level security restricts which rows are visible within an item the user can already open; it does not block opening the item itself.
Prepare Data
Question 4: An analyst wants to filter and aggregate data from a lakehouse's SQL analytics endpoint without writing any code. What should they use?
- A. The Visual Query Editor
- B. A Power BI Desktop Power Query step
- C. A KQL queryset
- D. A DAX measure
Why A is correct: The Visual Query Editor is Fabric's no-code option for filtering and aggregating data from a SQL analytics endpoint, which is exactly the tool the official DP-600 skills outline lists for this task.
Why B is wrong: Power Query/M is a PL-300-era Power BI Desktop tool and does not appear in the DP-600 skills outline.
Why C is wrong: KQL querysets are for Eventhouse/KQL data, not a lakehouse SQL analytics endpoint.
Why D is wrong: DAX operates inside a semantic model, not directly against a SQL analytics endpoint's raw data.
Medium
Questions 5–8Apply a concept to a scenario with one clear complicating factor.
Prepare Data
Question 5: A warehouse has a wide, denormalized sales table mixing customer, product, and transaction attributes in one place, making the semantic model hard to maintain. What should the analytics engineer implement?
- A. Leave the table as-is and add more DAX measures to compensate
- B. Convert the warehouse into a lakehouse
- C. Implement a star schema, splitting the table into fact and dimension tables
- D. Move all the data into an Eventhouse for better performance
Why C is correct: A star schema — separating transactional facts from descriptive dimensions like customer and product — is the documented pattern for exactly this maintainability and performance problem in a lakehouse or warehouse.
Why A is wrong: Adding more DAX measures treats a symptom, not the underlying data-modeling problem, and increases long-term maintenance burden.
Why B is wrong: Switching item type does not address table design; a lakehouse can have the same modeling problem a warehouse can.
Why D is wrong: An Eventhouse is for streaming/real-time data, not for restructuring an existing warehouse table for better modeling.
Implement and Manage Semantic Models
Question 6: A report needs its currency symbol and decimal precision to change automatically depending on which measure a user selects, without duplicating the visual for each currency. What DAX feature fits?
- A. A calculated column added to the fact table
- B. A new dimension table for currency
- C. A DirectQuery-only measure
- D. A dynamic format string tied to a calculation group or measure logic
Why D is correct: Dynamic format strings let a measure's displayed format change based on context or selection — exactly the behavior needed for a currency/precision display that adapts per selected measure.
Why A is wrong: A calculated column stores a static value per row; it does not change how a measure is formatted for display.
Why B is wrong: A currency dimension table could support filtering by currency, but it does not by itself change a measure's display format dynamically.
Why C is wrong: DirectQuery is a storage mode unrelated to how a value is formatted for display.
Maintain a Data Analytics Solution
Question 7: Before changing a shared lakehouse's schema, an analytics engineer wants to know every downstream warehouse, dataflow, and semantic model that depends on it. What should they run first?
- A. Impact analysis on the lakehouse's downstream dependencies
- B. A deployment pipeline promotion
- C. A full re-import of every connected semantic model
- D. A sensitivity label audit
Why A is correct: Impact analysis is specifically designed to surface downstream dependencies — warehouses, dataflows, semantic models — before a change is made, matching the exact requirement in the scenario.
Why B is wrong: A deployment pipeline promotes items between environments; it does not, by itself, enumerate downstream dependencies before a change.
Why C is wrong: Re-importing every model is disruptive and does not tell the engineer what depends on the lakehouse before making the change.
Why D is wrong: A sensitivity label audit addresses compliance classification, not structural dependency mapping.
Implement and Manage Semantic Models
Question 8: A semantic model needs to read lakehouse data directly, but the team also wants the row/column-level security defined in the SQL analytics endpoint's views to apply automatically. Which configuration fits?
- A. Direct Lake on OneLake
- B. Direct Lake on SQL analytics endpoint
- C. Import mode with a nightly refresh
- D. DirectQuery against the lakehouse files directly
Why B is correct: Direct Lake on SQL analytics endpoint routes through the endpoint's views and security layer, so security defined there applies — Direct Lake on OneLake bypasses that endpoint layer entirely.
Why A is wrong:
Why C is wrong: Import mode could technically carry the security if re-modeled, but it reintroduces a refresh cycle the team is trying to avoid and is not the documented fit for this requirement.
Why D is wrong: There is no supported DirectQuery mode against raw lakehouse files bypassing the SQL analytics endpoint in the way described.
Hard
Questions 9–12Diagnose a scenario with layered or non-obvious causes, the way real exam case studies do.
Implement and Manage Semantic Models
Question 9: A Direct Lake semantic model was fast at launch but has grown steadily slower over three months, with no visible errors. What is the most likely explanation to investigate first?
- A. Row-level security roles expired
- B. DAX always slows down as a model ages
- C. The workspace ran out of storage quota
- D. The model has silently fallen back to DirectQuery because a table now exceeds Direct Lake guardrails
Why D is correct: A gradual, error-free slowdown in a Direct Lake model is the classic symptom of a silent fallback to DirectQuery once a table grows past Direct Lake's guardrails — checking fallback status is the documented first diagnostic step.
Why A is wrong: RLS roles do not expire on a timer in a way that produces a gradual performance decline — this is a performance symptom, not an access-denial symptom.
Why B is wrong: DAX performance does not inherently degrade with model age on its own; something structural changed, which is what needs investigating.
Why C is wrong: Fabric workspace storage limits do not silently degrade query performance in this gradual way without a visible error.
Maintain a Data Analytics Solution
Question 10: A user has Viewer access at the workspace level, item-level access to a specific semantic model, but is also targeted by a row-level security role that filters out all rows for their region. What does this user see when they open a report on that model?
- A. The report shell with no rows visible for their region, since RLS applies on top of the access they already have
- B. An error, since the layers conflict
- C. The full dataset, since item-level access overrides row-level security
- D. Workspace-level Viewer access blocks them from opening the model at all
Why A is correct: Access-control layers stack rather than override each other: workspace and item-level access determine whether the user can open the item at all, while row-level security independently filters the rows they see once inside — so they see the report with their region's rows filtered out.
Why B is wrong: These are compatible, stacking layers by design, not conflicting rules that produce an error.
Why C is wrong: Item-level access controls whether the item can be opened; it does not override or bypass row-level security's row filtering, which is a separate, independent layer.
Why D is wrong: The scenario states the user has both workspace Viewer and item-level access, so they can open the model — RLS then filters the rows they see, it does not block the open action itself.
Prepare Data
Question 11: A source system delivers the same customer record multiple times with a mix of duplicate, missing, and null field values, and downstream reports need one clean row per customer before the star schema is built. What is the most appropriate place to resolve this?
- A. In DAX, using a calculated table on the semantic model
- B. In a SQL view or stored procedure during the transform step, before modeling
- C. By manually editing the source system's database
- D. By hiding the duplicate rows from report visuals only
Why B is correct: Resolving duplicates, missing values, and nulls is explicitly a transform-data objective on DP-600, meant to happen in a SQL view or stored procedure before the star schema and semantic model are built — cleaning data at the source of truth for every downstream consumer.
Why A is wrong: A DAX calculated table works inside the semantic model, after modeling has already happened, and would need to be redone in every model built on this data — the wrong layer for a shared, reusable fix.
Why C is wrong:
Why D is wrong: Hiding duplicates only in visuals leaves the underlying dirty data in place for every other report and consumer built on the same table.
Implement and Manage Semantic Models
Question 12: A very large semantic model refreshes nightly, but only a small fraction of rows change each day, and the nightly refresh window keeps growing. Which two optimizations most directly address this, together?
- A. Switch to DirectQuery and disable refresh entirely
- B. Add more calculated columns to speed up queries
- C. Implement incremental refresh, and evaluate large semantic model storage format if the model is approaching size limits
- D. Move the model to a different workspace
Why C is correct: Incremental refresh directly targets refreshing only changed partitions instead of the whole model, and large semantic model storage format is the documented option for models approaching default size limits — together they address both symptoms in the scenario.
Why A is wrong: Disabling refresh does not solve a growing refresh-window problem — it removes freshness entirely rather than optimizing the process.
Why B is wrong: Calculated columns add to model size and processing time; they do not address refresh duration and can make the problem worse.
Why D is wrong: Moving workspaces does not change how much data is reprocessed on refresh — it does not address the root cause.
Score Yourself: The Answer Key
| Question | Tier | Correct Answer |
|---|---|---|
| Question 1 | Easy | B |
| Question 2 | Easy | C |
| Question 3 | Easy | D |
| Question 4 | Easy | A |
| Question 5 | Medium | C |
| Question 6 | Medium | D |
| Question 7 | Medium | A |
| Question 8 | Medium | B |
| Question 9 | Hard | D |
| Question 10 | Hard | A |
| Question 11 | Hard | B |
| Question 12 | Hard | C |
Common Questions About This DP-600 Set
Are these real DP-600 exam questions?
No. Microsoft does not release its actual exam questions. These are original scenario-based questions we wrote against the official DP-600 skills outline to match its style and depth.
Why organize by difficulty instead of by domain?
Domain-grouped samples are common; this set is organized by difficulty instead so you can gauge where your actual skill level sits before committing to a full domain-by-domain study plan.
How many questions does the full DP-600 bank have?
MSCertQuiz maintains 500 DP-600 questions across all three domains, weighted to match the official domain percentages. Forty are free; the twelve above are a sample of that free set.
Should I take these before or after reading the DP-600 study guide?
Either order works, but if a term above is unfamiliar, the DP-600 study guide covers all three domains plus how DP-600 compares to PL-300 and DP-700.
Why do the Hard questions feel like they combine multiple concepts?
That mirrors DP-600 itself — Microsoft's audience profile describes candidates managing analytics assets end-to-end, and real scenario questions often require ruling out several plausible-looking causes, not recalling one isolated fact.
Is DP-600 harder than PL-300?
DP-600 assumes and extends PL-300-level DAX and modeling knowledge, then adds governance, deployment pipelines, and SQL/KQL-based data preparation that PL-300 does not test — so most PL-300 holders find DP-600 broader rather than simply "harder."
MSCertQuiz sells practice-exam access for DP-600, and these questions were written by the same team that maintains the 500-question bank. Domain names above trace to Microsoft Learn's official DP-600 study guide, checked September 7, 2026. For the reasoning behind each domain, see the DP-600 study guide; for a dense task-organized reference, see the DP-600 cheat sheet.
Want the Other 488 DP-600 Questions?
Take a full timed DP-600 readiness quiz across all three domains and see your estimated readiness before exam day.
Take the Full DP-600 Mock Exam →