Northwind Retail needs business analysts to build a currency-conversion transform themselves without writing code — which Fabric item fits?

Dataflow Gen2 — the low-code, Power Query-style item designed for self-service transforms without custom code.

Northwind Retail wants schema changes reviewed in Test before they reach the production workspace — what enforces that?

A deployment pipeline, which stages and promotes items through Dev/Test/Prod rather than just tracking source changes.

Northwind must let support staff see customer records but not raw card numbers, without altering the stored data — which control fits?

Dynamic data masking, which hides column values for unauthorized users without changing underlying data.

A new Northwind contractor should see only one report in the workspace, not every item in it — what access level enforces that?

Item-level access control, granted on the single report rather than the whole workspace.

Contoso Logistics has a 500-million-row shipment fact table where 99% of rows never change — how should the nightly load be designed?

An incremental load using a watermark column, avoiding a full reload of unchanged historical data.

Contoso needs a continuously synced, near real-time copy of an operational Azure SQL Database in OneLake without building a pipeline — which feature fits?

Mirroring, which continuously replicates a source database into OneLake, unlike a shortcut which only references existing data.

Contoso’s Python-first team needs branching, custom transformation logic over large Parquet files in a Lakehouse — which tool fits?

A notebook using PySpark, which supports custom code and conditional logic that Dataflow Gen2 and plain T-SQL do not.

Contoso’s nightly shipment-status rollup looks wrong because some status events arrive after the rollup already ran — what design fixes it?

Extend the incremental load with a grace-period window so late-arriving records are captured on the next run instead of being dropped or forcing full reloads.

Fabrikam Health’s Power BI reports are stale because a semantic model refresh silently failed overnight — what should be configured?

An alert on semantic model refresh monitoring, so failures surface immediately instead of being discovered by end users.

A Fabrikam Spark job fails with an out-of-memory error — where should you look first to diagnose it?

The notebook’s Spark application run details/logs, not the pipeline monitoring hub, since the failure is inside the Spark job itself.

A Fabrikam Lakehouse table has thousands of small files and reads have gotten slow even though the Spark pool is appropriately sized — what fixes it?

Run table maintenance (OPTIMIZE with V-Order) to compact small files — a storage-layer fix, not a compute-layer one like resizing the Spark pool.

A Fabrikam OneLake shortcut intermittently shows missing rows that exist in the source system — what should be checked first?

The shortcut’s target configuration and permissions, since a broken or misconfigured shortcut is the most common cause of this exact symptom.

Free DP-700 Practice Questions (2026): Fabric Scenarios

TL;DR:DP-700 rarely asks "what is a Lakehouse" — it describes a situation and asks which Fabric tool or setting fits it. These 14 questions are grouped into three extended scenarios, one per exam domain, so you practice the same reading-comprehension pattern the real exam uses instead of isolated trivia.

MSCertQuiz sells a 500-question DP-700 bank; these 14 are free and written by the same team, sourced against Microsoft's DP-700 skills-measured objectives (as of July 21, 2026). For the full domain breakdown behind these scenarios, see the DP-700 study guide.

How These Scenarios Work

Each scenario describes one fictional company's Fabric setup, then asks 4–5 questions against that same setup — exactly how DP-700's own case-study-style items work. Pick an answer before reading the rationale below each question; the value is in seeing why the three wrong options fail, not just which letter is correct.

Scenario 1: Northwind Retail — Implementing and Managing the Analytics Solution

Northwind Retail runs its e-commerce reporting in a single Fabric workspace. The data platform team is setting up governance before onboarding more business users: who can build what, how changes reach production, and who can see which data.

Self-service currency conversion

Northwind’s finance analysts want to build and adjust a currency-conversion transform themselves, without writing code, and without waiting on the data engineering team.

Which Fabric item should the platform team point them to?

A) Dataflow Gen2

B) A notebook using PySpark

C) A pipeline Copy activity

D) A KQL queryset

✓ Correct answer: A. Dataflow Gen2 is Fabric’s low-code, Power Query-style transform item, purpose-built so non-engineers can build and adjust transforms visually.

B) Notebooks require writing PySpark code — the opposite of what a self-service, no-code requirement calls for.

C) A Copy activity moves data between locations; it doesn’t provide a transform-building interface for business users.

D) KQL querysets are for querying Eventhouse data, not for building reusable transform logic over batch data.

Reviewed promotion to production

The team wants every schema or logic change reviewed in a Test workspace before it can reach the Production workspace — an approval gate, not just a change history.

What should they configure?

A) Git integration only

B) A deployment pipeline

C) Workspace-level access control

D) A sensitivity label

✓ Correct answer: B. Deployment pipelines are Fabric’s mechanism for staging and promoting items through Dev/Test/Prod, which is exactly the reviewed-promotion workflow described.

A) Git integration tracks and stores source changes; it doesn’t stage or gate promotion between environments by itself.

C) Workspace-level access controls who can enter a workspace, not how changes move between workspaces.

D) Sensitivity labels classify data (e.g. Confidential); they have nothing to do with release promotion.

Shortcut returns no rows

An analyst can query a Warehouse table directly and sees all rows, but a OneLake shortcut pointing at the same underlying files returns zero rows for that same analyst.

What is most likely misconfigured?

A) Row-level security on the Warehouse

B) The analyst’s workspace role

C) OneLake security on the shortcut’s target

D) A sensitivity label on the report

✓ Correct answer: C. OneLake security governs access to the underlying files independently of which Fabric item reads them — a shortcut inherits that permission boundary even when the Warehouse item itself is accessible.

A) Row-level security would restrict specific rows in both access paths consistently, not zero out only the shortcut.

B) A workspace role problem would typically block the Warehouse too, not just the shortcut.

D) Sensitivity labels classify data for compliance; they don’t block query results.

Masking card numbers

Support staff need to see customer order records to resolve tickets, but must never see full card numbers — and the underlying stored data must stay unchanged for the finance team, who do need full visibility.

Which control fits?

A) OneLake security

B) Row-level security

C) A sensitivity label

D) Dynamic data masking

✓ Correct answer: D. Dynamic data masking hides specific column values for unauthorized users at query time without altering the underlying stored data, letting finance keep full visibility.

A) OneLake security controls file/folder-level access, not column-level value masking.

B) Row-level security restricts which rows a user sees, not which column values within a visible row.

C) Sensitivity labels classify and flag data; they don’t hide values from specific roles.

Contractor scoped to one report

A new contractor needs to view exactly one sales report and nothing else in a workspace that also contains several other reports, a Lakehouse, and two pipelines.

What should be granted?

A) Item-level access on just that report

B) Workspace-level Viewer access

C) A OneLake security rule

D) An audit log exemption

✓ Correct answer: A. Item-level access control grants permission on a single item (the report) without exposing the rest of the workspace’s contents.

B) Workspace-level access would expose every item in the workspace, not just the one report.

C) OneLake security governs underlying file access, not visibility of a Power BI report item.

D) Audit log exemptions don’t exist as an access mechanism and wouldn’t scope visibility anyway.

Scenario 2: Contoso Logistics — Ingesting and Transforming Data

Contoso Logistics tracks shipments across a batch operational database and a live stream of scanner events from warehouse floors. The data engineering team is designing how both feed into Fabric.

A 500-million-row fact table

Contoso’s shipment fact table has 500 million rows. On a typical night, fewer than 0.1% of rows are new or changed; the rest are historical and untouched.

How should the nightly load be designed?

A) Full load, truncate and reload every night

B) Incremental load using a watermark column

C) Mirroring from the source database

D) A OneLake shortcut to the source

✓ Correct answer: B. An incremental load using a watermark (e.g. a last-modified timestamp) only pulls new or changed rows, avoiding the cost of reloading 500 million mostly-unchanged rows every night.

A) A full reload of 500M rows nightly wastes compute and time for a dataset that’s 99.9% unchanged.

C) Mirroring is for continuous replication of an operational database, not a scheduled analytical fact table load.

D) A shortcut references data in place; it doesn’t implement incremental load logic for a data warehouse-style load.

Near real-time sync, no pipeline

Contoso wants a continuously up-to-date copy of its operational Azure SQL Database available in OneLake for reporting, and wants to avoid building and maintaining a scheduled copy pipeline.

What should they use?

A) A pipeline Copy activity on a 5-minute schedule

B) A OneLake shortcut

C) Mirroring

D) Dataflow Gen2 with a scheduled refresh

✓ Correct answer: C. Mirroring continuously replicates a source database into OneLake in near real time without a maintained pipeline — a purpose-built feature for exactly this requirement.

A) A scheduled Copy pipeline is exactly the maintained pipeline the requirement says to avoid, and isn’t truly near real-time.

B) A shortcut only references existing data at its source location; it doesn’t replicate an operational database into OneLake.

D) Dataflow Gen2 on a schedule has the same maintenance and latency drawbacks as a scheduled pipeline.

Custom branching logic over Parquet

Contoso’s data engineers are Python-first and need to apply custom, conditional transformation logic (different handling per shipment carrier) over large Parquet files sitting in a Lakehouse.

Which tool fits best?

A) T-SQL in a Warehouse

B) Dataflow Gen2

C) A KQL queryset

D) A notebook using PySpark

✓ Correct answer: D. Notebooks with PySpark support custom code and conditional branching logic at scale over Lakehouse files, matching both the team’s skill set and the requirement.

A) T-SQL against a Warehouse works for relational, set-based logic, but doesn’t match Parquet-in-Lakehouse plus a Python-first team.

B) Dataflow Gen2’s low-code interface is not built for complex custom conditional branching by carrier.

C) KQL querysets query Eventhouse/streaming data, not batch Parquet files in a Lakehouse.

No-code event filtering

Contoso’s ops team (non-developers) wants to filter out test-scanner noise and route real shipment-scan events before they land in the Eventhouse, without writing code.

What handles this?

A) Eventstreams

B) Spark structured streaming in a notebook

C) A KQL update policy

D) A pipeline

✓ Correct answer: A. Eventstreams is Fabric’s no-code item for routing and lightly transforming streaming data in flight, matching the non-developer, no-code requirement exactly.

B) Spark structured streaming requires writing code in a notebook, which the ops team explicitly can’t do.

C) A KQL update policy runs inside the Eventhouse after ingestion, not as a pre-ingestion no-code filter.

D) Pipelines orchestrate scheduled/batch activities; they aren’t built for continuous event-stream filtering.

Late-arriving status updates

Contoso’s nightly shipment-status rollup sometimes shows incomplete totals because a small number of status events for "yesterday" arrive a few hours after the rollup already ran.

What design change addresses this?

A) Switch permanently to a full reload every run

B) Extend the incremental load with a grace-period window that reprocesses recent data on the next run

C) Ignore late records — they’re within acceptable error

D) Increase the pipeline’s timeout setting

✓ Correct answer: B. A grace-period window re-includes a recent slice of data on the next run, catching records that arrived after the prior run without reprocessing the entire dataset.

A) A full reload every run fixes the symptom but at a cost disproportionate to a small volume of late records.

C) Silently ignoring late-arriving data produces the exact incomplete totals the scenario describes as a problem.

D) A longer pipeline timeout addresses slow-running jobs, not data that hadn’t been generated yet when the job ran.

Scenario 3: Fabrikam Health — Monitoring and Optimizing the Analytics Solution

Fabrikam Health runs a Fabric-based reporting platform for clinic operations. The platform has been live for months, and the team is now dealing with the operational reality of keeping it healthy and fast.

Silent refresh failure

A semantic model feeding Fabrikam’s daily operations dashboard failed to refresh overnight, and nobody knew until a clinic manager noticed stale numbers the next morning.

What should the team configure to prevent this recurring?

A) A deployment pipeline

B) Row-level security on the semantic model

C) An alert on semantic model refresh monitoring

D) A OneLake shortcut

✓ Correct answer: C. Configuring an alert on semantic model refresh monitoring surfaces failures immediately to the team instead of end users discovering stale data first.

A) Deployment pipelines manage promoting changes between environments, not runtime refresh failures.

B) Row-level security controls who sees which rows; it has no relationship to refresh monitoring.

D) A shortcut is a data-access mechanism, unrelated to refresh scheduling or failure alerting.

Out-of-memory Spark job

A nightly notebook transforming patient-visit data fails with an out-of-memory error partway through, and the on-call engineer needs to find out why fast.

Where should they look first?

A) The pipeline monitoring hub run history

B) The Warehouse query history

C) OneLake audit logs

D) The notebook’s Spark application run details/logs

✓ Correct answer: D. The failure originates inside the Spark job itself, so the notebook’s Spark application logs contain the actual memory/executor details needed to diagnose it.

A) The pipeline monitoring hub shows whether an activity succeeded or failed, but not the underlying Spark memory diagnostics.

B) Warehouse query history covers T-SQL queries against the Warehouse, not a notebook’s Spark execution.

C) OneLake audit logs record access events, not Spark job execution failures.

Slow reads despite a sized-right cluster

Reads against a Fabrikam Lakehouse table have become noticeably slower over several months. The Spark pool is already sized appropriately for the workload, and the table has accumulated thousands of small files from frequent small writes.

What should the team do?

A) Run table maintenance (OPTIMIZE with V-Order) to compact the small files

B) Increase the Spark pool size further

C) Move the table to a Warehouse instead

D) Disable OneLake security on the table

✓ Correct answer: A. Many small files is a storage-layer problem; compacting them via OPTIMIZE/V-Order directly addresses the read-performance symptom described.

B) The pool is already stated as appropriately sized — this is a storage-layer issue, not a compute-capacity one.

C) Moving engines doesn’t fix an underlying small-file problem and introduces unnecessary migration risk.

D) Security settings have no effect on file layout or read performance.

Intermittently missing shortcut rows

A OneLake shortcut Fabrikam built to an external clinical records store intermittently returns fewer rows than the source system actually has, with no pattern by time of day.

What should be checked first?

A) The Spark pool’s autoscale settings

B) The shortcut’s target configuration and permissions

C) The Warehouse’s statistics

D) The workspace’s domain settings

✓ Correct answer: B. Intermittent, unexplained missing rows through a shortcut most commonly trace back to a misconfigured target path or a permissions gap at the source, which is exactly what OneLake shortcut error troubleshooting targets first.

A) Spark autoscale affects compute for Spark jobs, not shortcut result completeness.

C) Warehouse statistics affect T-SQL query planning, unrelated to a shortcut pointing elsewhere.

D) Domain workspace settings group workspaces administratively; they don’t affect a specific shortcut’s data completeness.

Distractor Patterns at a Glance

Across these 14 questions, wrong answers cluster into a few repeatable traps. Recognizing the trap type is often faster than re-deriving the right tool from scratch.

Trap pattern	Seen in	How to spot it
Right category, wrong layer	Q3, Q13, Q14	A distractor is a real Fabric feature, just operating at the wrong layer (compute vs storage, item vs file-level).
Correct tool, wrong maintenance tradeoff	Q6, Q7	The distractor technically works but reintroduces the exact cost/maintenance burden the scenario says to avoid.
No-code vs code mismatch	Q1, Q8, Q9	The scenario states who is doing the work (business user vs engineer); the distractor requires the skill set the persona doesn't have.
Symptom-matching, not cause-matching	Q10, Q12	Distractors treat the visible symptom (slow job, incomplete numbers) rather than its root cause.

Ready for a Full Timed DP-700 Mock Exam?

Take a free, timed readiness quiz built from the same domain weightings covered here.

Take the DP-700 Readiness Quiz →

DP-700 Practice Questions FAQ

Are these questions the same difficulty as the real DP-700 exam?

They're built to match DP-700's scenario-based format and the three domains Microsoft publishes, but no practice source can reproduce Microsoft's exact item bank. Treat a strong score here as evidence you understand the tool-selection logic, not a guarantee of a specific exam score.

Why are the questions grouped into scenarios instead of by domain heading?

DP-700's real exam frequently poses several questions against one extended case study, testing whether you can hold context across questions the way you would on a real project. Practicing that format matters as much as practicing the content.

How many questions does the full DP-700 practice bank have?

MSCertQuiz's DP-700 bank has 500 questions for paid access; the 14 here are free and cover all three domains at roughly the same weighting as the real exam.

Do I need Fabric access to make sense of these questions?

No, but it helps. If a term like "OneLake shortcut" or "deployment pipeline" feels unfamiliar, the DP-700 cheat sheet defines every term used across these scenarios.

What if I got several questions wrong in one scenario?

Re-read that scenario's intro paragraph first — most wrong answers trace back to missing a constraint stated in the setup (who's doing the work, what they explicitly want to avoid), not a knowledge gap about the Fabric feature itself.

Are these questions updated when Microsoft changes the DP-700 objectives?

MSCertQuiz tracks Microsoft's published skills-measured changes and updates the underlying question bank accordingly; these free samples reflect objectives as of July 21, 2026.

Free DP-700 Practice Questions: 3 Fabric Scenarios, 14 Questions