Microsoft FabricJuly 7, 202612 min read

DP-700 Practice Questions (2026): 10 Free Microsoft Fabric Data Engineer Questions with Answers

Ten exam-style questions across all three DP-700 domains — each with the correct answer and a plain-English explanation on click. Score yourself at the end to see if you are exam-ready.

40–60Questions
100 minDuration
700/1000To pass
3Domains
$165USD fee

The DP-700: Implementing Data Engineering Solutions Using Microsoft Fabric exam is the hands-on gate for the Microsoft Certified: Fabric Data Engineer Associate badge. It rewards people who have actually built pipelines, lakehouses, and warehouses in Fabric — wiring up OneLake shortcuts, medallion layers, Dataflows Gen2, Spark notebooks, T-SQL, and KQL — not people who only memorised feature names.

Below are 10 free DP-700 practice questions written in the same style as the real exam, spread evenly across the three domains of the 2026 blueprint. Read each one, pick your answer, then hit Show answer & explanation. Keep score — there is a readiness check at the bottom.

The 10 questions Domain coverage Score yourself FAQ
How to use this: answer honestly before revealing. The explanation matters more than the letter — if you got it right for the wrong reason, treat it as wrong and revisit that feature in the Microsoft Fabric docs.

10 DP-700 practice questions

Q1EasyImplement & manage

A team of SQL developers needs multi-table ACID transactions, stored procedures, and full T-SQL INSERT/UPDATE/DELETE over structured data for BI reporting. Which Fabric item should you choose?

  1. A Lakehouse, queried through its SQL analytics endpoint
  2. A Warehouse
  3. An Eventhouse (KQL database)
  4. A Dataflow Gen2
Show answer & explanation
Correct: B

A Fabric Warehouse gives full T-SQL DQL, DML, and DDL plus multi-table ACID transactions and stored procedures — ideal for SQL-first teams. The Lakehouse SQL analytics endpoint (A) is read-only: it supports queries and views but no INSERT/UPDATE/DELETE. An Eventhouse (C) is queried with KQL for real-time telemetry, not T-SQL DML. A Dataflow Gen2 (D) transforms data; it is not a queryable relational store.

Q2MediumImplement & manage

A data engineer must create and edit notebooks and pipelines in a workspace, but must not be able to share items or add other users. Which workspace role grants the least privilege that meets this?

  1. Viewer
  2. Contributor
  3. Member
  4. Admin
Show answer & explanation
Correct: B

Contributor can view and modify all content (create/edit notebooks, pipelines, lakehouses) but cannot reshare items or manage access. Viewer (A) is read-only and cannot create items. Member (C) adds sharing and can add lower-privilege users. Admin (D) can manage permissions and delete the workspace — both exceed what is asked.

Q3MediumImplement & manage

A citizen developer must ingest data from a SaaS source and apply low-code, Power Query-style transformations (merge, unpivot, change column types) with no code. Which Fabric item fits best?

  1. A Spark notebook written in PySpark
  2. A Dataflow Gen2
  3. A data pipeline with only a Copy activity
  4. A KQL queryset
Show answer & explanation
Correct: B

Dataflow Gen2 is Fabric's low-code, Power Query transformation engine — perfect for merge/unpivot/type changes without code. A notebook (A) is code-first (PySpark/SQL). A Copy activity (C) moves data but does not perform rich row-level transforms. A KQL queryset (D) queries real-time data in an Eventhouse, not SaaS batch loads.

Q4HardImplement & manage

In a Fabric Warehouse, each regional manager must query the Sales table but see only rows for their own region, while everyone sees an obfuscated version of the customer email column. Which combination meets both requirements?

  1. Column-level security plus object-level security
  2. Row-level security (RLS) plus dynamic data masking
  3. Sensitivity labels plus the workspace Viewer role
  4. OneLake security plus folder-level access
Show answer & explanation
Correct: B

Row-level security applies a predicate so each manager sees only their region's rows, and dynamic data masking obfuscates a column's values (email) at query time without changing stored data. Column-level security (A) hides an entire column rather than masking part of it, and object-level security hides whole tables. Sensitivity labels (C) classify data but do not filter rows. Folder/file access (D) governs OneLake files, not warehouse row/column visibility.

Q5EasyIngest & transform

Your bronze layer needs live access to Delta files that already sit in an Amazon S3 bucket — no duplication, no ETL pipeline to maintain, and always in sync with the source. What should you create?

  1. A data pipeline Copy activity that imports the files nightly
  2. A OneLake shortcut to the S3 location
  3. A Dataflow Gen2 that appends the files on a schedule
  4. A mirrored database
Show answer & explanation
Correct: B

A OneLake shortcut references external data (S3, ADLS Gen2, another lakehouse) in place, with no copy and live access — exactly the medallion recommendation for the bronze layer. Copy (A) and Dataflow (C) duplicate the data and add a job to maintain. Mirroring (D) replicates operational databases such as Azure SQL, Cosmos DB, or Snowflake — not raw Delta files in a storage bucket.

Q6MediumIngest & transform

A pipeline reloads an entire 500-million-row source table every night, which is slow and expensive. You want to load only the rows that changed since the last run. Which loading pattern should you implement?

  1. A full load that overwrites the target every night
  2. An incremental load using a watermark column (for example, LastModified) to fetch only new or changed rows
  3. Increase the Spark pool size so the full load finishes faster
  4. Enable V-Order on the target table
Show answer & explanation
Correct: B

An incremental load tracks a high-watermark (such as the maximum LastModified value from the previous run) and pulls only rows newer than it, sharply cutting the volume moved. A full overwrite (A) is the problem being solved. A bigger Spark pool (C) processes the same huge volume faster but still reads everything. V-Order (D) is a read optimization and does nothing to reduce what is ingested.

Q7HardIngest & transform

You must process a continuous stream of IoT telemetry, compute 5-minute tumbling-window aggregates, and land the results for real-time dashboards with sub-second query latency. Which approach fits best?

  1. A Dataflow Gen2 that writes to a Lakehouse table
  2. A nightly data pipeline that loads a Warehouse
  3. An Eventstream into an Eventhouse (KQL database), aggregated with KQL windowing functions
  4. A Warehouse stored procedure scheduled to run every minute
Show answer & explanation
Correct: C

This is Real-Time Intelligence: an Eventstream ingests and routes the stream, an Eventhouse / KQL database stores high-volume event data for sub-second queries, and KQL windowing functions compute tumbling-window aggregates. Dataflow Gen2 (A) and nightly pipelines (B) are batch, not streaming. A per-minute stored procedure (D) is micro-batch polling that cannot deliver true streaming latency or windowing.

Q8EasyMonitor & optimize

You need one place to see the run history, status, and durations of pipeline runs, notebook runs, and Dataflow Gen2 refreshes across a workspace. Where do you look?

  1. The Monitoring hub
  2. The Lakehouse explorer
  3. The Git integration panel
  4. The Fabric Capacity Metrics app
Show answer & explanation
Correct: A

The Monitoring hub is the central view of run history and status for Fabric items — pipelines, notebooks, dataflows, and semantic model refreshes — with durations and error drill-down. The Lakehouse explorer (B) browses data, not runs. The Git panel (C) tracks source-control state. The Capacity Metrics app (D) reports capacity utilisation and throttling, not individual item run history.

Q9MediumMonitor & optimize

A Lakehouse Delta table is fed by frequent micro-batch writes and has accumulated thousands of tiny Parquet files, which is slowing reads. Which maintenance command consolidates them into fewer, larger files?

  1. VACUUM
  2. OPTIMIZE (bin compaction)
  3. ANALYZE TABLE
  4. REFRESH TABLE
Show answer & explanation
Correct: B

OPTIMIZE performs bin compaction, merging many small files into fewer larger ones to cut file-management overhead and speed scans. VACUUM (A) removes old unreferenced files to reclaim storage but does not compact active data. ANALYZE/REFRESH (C, D) do not rewrite file layout. Note that OPTIMIZE is a Spark SQL command — run it in a notebook, Spark job, or the Lakehouse maintenance UI, not in the SQL analytics endpoint.

Q10HardMonitor & optimize

A gold Lakehouse table feeds a Power BI Direct Lake semantic model, but cold-cache report queries are slow. You want a write-time optimization that reorganises the Parquet layout to speed those reads. What should you apply?

  1. Run VACUUM with a zero-hour retention
  2. Apply V-Order to the table (for example, OPTIMIZE … VORDER)
  3. Switch the report from Direct Lake to DirectQuery
  4. Disable auto compaction on the table
Show answer & explanation
Correct: B

V-Order is a write-time Parquet optimization (reordered layout, encoding, compression) that markedly improves read performance for Power BI Direct Lake and the SQL analytics endpoint — especially cold-cache queries. VACUUM (A) only reclaims storage. DirectQuery (C) adds query latency versus Direct Lake. Disabling auto compaction (D) leaves small files and makes reads worse.

What these questions cover

The 10 questions are spread across the real DP-700 blueprint, so your score here is a rough proxy for the live exam. These are the three domains and their approximate exam weights — each carries roughly a third of the marks:

Implement and manage an analytics solution30–35%
Ingest and transform data30–35%
Monitor and optimize an analytics solution30–35%

Score yourself

Count how many of the 10 you got right before revealing the answer. Then read the band you land in honestly — the goal is a real pass, not a good feeling.

0–5Not ready yet. You are guessing on core Fabric concepts. Rebuild fundamentals — lakehouse vs warehouse, shortcuts, medallion, and Delta basics — before booking.
6–8Close. You know the items but slip on edge cases (RLS vs masking, incremental loads, OPTIMIZE vs V-Order). Drill a full timed bank and target weak domains.
9–10Exam-ready signal. Confirm with a couple of full-length timed mocks under exam conditions, then book with confidence.

Want the full DP-700 question bank?

These 10 are a taster. ExamCert runs hundreds of exam-style DP-700 questions with explanations, timed mocks, and weak-domain tracking — the real exam interface, on your phone.

Free practice test DP-700 exam guide

DP-700 exam FAQ

How many questions are on the DP-700 exam?

Microsoft does not publish a fixed count, but candidates typically see 40 to 60 questions in about 100 minutes, mixing multiple choice, multi-select, drag-and-drop, and hotspot items, and sometimes a Fabric case study. Read any case-study scenario fully before answering, since those items carry extra weight.

What score do I need to pass DP-700?

A scaled 700 out of 1000. Because scoring is scaled, 700 is not the same as 70% of questions correct — harder and case-study items carry more weight toward the total.

What does DP-700 cover in 2026?

Three equally weighted domains (about a third each): implement and manage an analytics solution, ingest and transform data, and monitor and optimize an analytics solution. Expect lakehouse vs warehouse decisions, OneLake shortcuts, medallion architecture, Dataflows Gen2, pipelines, Spark/PySpark, T-SQL, KQL and Real-Time Intelligence, security, and Delta table optimization.

Are practice questions enough to pass?

They are essential but not sufficient. DP-700 is hands-on — build a lakehouse and a warehouse in a Fabric trial capacity, load data with pipelines and notebooks, query with T-SQL and KQL, and use practice questions to find and close weak domains.

ExamCert Team — we build exam-style practice banks and prep apps for 90+ IT certifications. Questions here are original, written to match the DP-700 objectives; they are not real exam items.

Related: DP-700 exam guide · DP-700 complete study guide · Free practice tests