DP-900 Study Guide 2026: Pass Azure Data Fundamentals

DP-900 is the exam behind Microsoft Certified: Azure Data Fundamentals — a beginner-level credential that proves you understand core data concepts and how they map to Azure's data services. No coding. No prior cloud experience. No labs. Just a solid grasp of what structured data is, when to reach for a relational database versus a NoSQL store, and which Azure service does what.

That makes DP-900 one of the cleanest "first data cert" wins out there: $99, a single 45-minute sitting, and a credential that does not expire. This guide covers exactly what's tested, the topics that show up over and over, memory aids to lock them in, and a one-week plan to get you through.

If you're still deciding whether to sit it at all, read Is DP-900 Worth It? first. If you're already committed, keep going.

What DP-900 actually is

DP-900 validates foundational data knowledge in an Azure context. Microsoft designed it for people entering data or cloud careers — students, career-changers, and non-technical stakeholders like project managers or pre-sales staff who need to speak the language without writing the queries.

You'll learn to describe:

• Core data concepts — structured, semi-structured, and unstructured data; transactional vs analytical workloads; the data roles on a team
• Relational data on Azure — the Azure SQL family and the open-source database options
• Non-relational data on Azure — Blob, Files, Table storage, and Cosmos DB
• Analytics workloads on Azure — ingestion, processing, Microsoft Fabric, Databricks, and Power BI

It is a fundamentals exam. The goal is recognition and mapping, not implementation.

Exam structure

• ~40–60 questions (typically around 40–50)
• 45 minutes
• Passing score: 700 on a 1–1000 scale (it's a scaled score, not a raw percentage)
• Mostly multiple-choice and drag-and-drop matching
• No labs, no coding
• Cost: ~$99 USD
• Prerequisites: none
• Validity: does not expire — Fundamentals certs have no renewal requirement

That last point matters. Role-based Azure certs (like DP-203 or Azure Administrator) expire and need annual renewal. DP-900, as a Fundamentals cert, is yours permanently once you pass.

The four skills-measured domains

Microsoft publishes the exact weightings. Study to them.

Domains 1 and 4 are the heavy hitters at up to 30% each. Together they can account for well over half the exam. Non-relational (Domain 3) is the lightest at 15–20% — but it's dense with "which storage type is this?" questions, so don't skip it.

Domain 1 — Core data concepts (25–30%)

• Structured, semi-structured, and unstructured data — and how to tell them apart
• Common file formats and data stores
• Transactional vs analytical workloads
• Data roles: database administrator (DBA), data engineer, data analyst

Domain 2 — Relational data on Azure (20–25%)

• Relational concepts, normalization, SQL, and database objects (tables, views, indexes, stored procedures)
• The Azure SQL family
• Azure Database for MySQL and Azure Database for PostgreSQL (managed open-source options)

Domain 3 — Non-relational data on Azure (15–20%)

• Azure Blob, Azure Files, and Azure Table storage
• Azure Cosmos DB — use cases and its multiple APIs

Domain 4 — Analytics workloads on Azure (25–30%)

• Data ingestion and processing
• Azure Databricks and Microsoft Fabric
• Batch vs streaming
• Power BI for visualization

The highest-yield topics

Certain themes generate a disproportionate share of the questions. Master these and you clear the passing line comfortably.

Data type identification

The exam loves to hand you an example and ask "which type is this?"

• Structured = fits neat rows and columns (a SQL table of customers)
• Semi-structured = has tags or keys but a flexible shape (JSON, XML)
• Unstructured = no predefined model (images, video, PDFs, audio)

Expect several of these. They're free points once the definitions are reflexive.

Transactional vs analytical (OLTP vs OLAP)

• OLTP (transactional) = many small, fast writes — think order processing, a point-of-sale system
• OLAP (analytical) = fewer, large reads aggregated for reporting and dashboards

If a scenario is about recording day-to-day operations, it's OLTP. If it's about analyzing history to make decisions, it's OLAP.

The Azure SQL family

This trips people up because there are three options with different trade-offs:

• Azure SQL Database — a single managed database, fully PaaS. Least admin overhead.
• Azure SQL Managed Instance — near-100% SQL Server compatibility, still PaaS. The migration bridge for existing SQL Server workloads.
• SQL Server on Azure VM — full SQL Server running on a VM, which is IaaS. Maximum control, maximum admin.

Know which is PaaS vs IaaS and why you'd pick each.

Azure storage types

Three names, three jobs:

• Blob — unstructured binary large objects (files, images, backups). Has access tiers: hot, cool, archive.
• Files — managed SMB file shares you can mount like a network drive
• Table — a NoSQL key-value store for tabular data

Cosmos DB is multi-model

This is a favorite question. Cosmos DB is multi-model — it supports several APIs so you can use the model that fits: NoSQL (SQL), MongoDB, Cassandra, Gremlin, and Table. If a question implies Cosmos DB only does one thing, it's wrong.

The analytics stack

• Microsoft Fabric — Microsoft's unified analytics platform (the 2026 exam leans into this)
• Azure Databricks — Apache Spark-based analytics and data engineering
• Azure Synapse — enterprise analytics/data warehousing
• Batch vs streaming — really a question about bounded (finite, processed in chunks) vs unbounded/continuous data
• Power BI — the visualization and reporting layer

Memory aids

Small hooks that stick:

• "S-M-V" for the Azure SQL family — SQL Database (single, PaaS) → Managed Instance (the bridge, PaaS) → VM (full control, IaaS). Read left to right: left = least admin, right = most control.
• Storage trio — "Blob = Binary large objects, Table = Tabular NoSQL, Files = File shares."
• OLTP vs OLAP — "OLTP = many small writes (orders), OLAP = few big reads (reports)."
• Normalization — it reduces redundancy. Know why (less duplication, fewer update anomalies), not "it's always better."
• Batch vs streaming — bounded vs continuous. Batch has a start and end; streaming never stops.

Common gotchas

• Managed Instance vs SQL Database vs SQL Server on VM. The exam wants the PaaS-vs-IaaS trade-off, not just the names.
• Blob vs Table vs Files. Match the storage type to the data shape — objects, key-value, or file shares.
• Normalization is not automatically "good." It's a trade-off that reduces redundancy; over-normalizing can hurt read performance.
• Batch vs streaming is a data-boundedness question in disguise.
• Cosmos DB is multi-model — don't fall for answers that pin it to a single API.

A one-week study plan

Assumes ~1.5–2 hours a day. Stretch it to two weeks if you're brand new to data concepts.

• Day 1 — Core data concepts. Structured/semi/unstructured, file formats, OLTP vs OLAP, the data roles. This is up to 30% of the exam; get it solid.
• Day 2 — Relational data. Relational concepts, normalization, SQL and database objects, and the Azure SQL family (drill S-M-V).
• Day 3 — Managed open-source + review. Azure Database for MySQL/PostgreSQL. Re-drill PaaS vs IaaS across the SQL family.
• Day 4 — Non-relational data. Blob (and its hot/cool/archive tiers), Files, Table, and Cosmos DB's APIs. Lock in that Cosmos DB is multi-model.
• Day 5 — Analytics workloads. Ingestion and processing, Fabric, Databricks, Synapse, batch vs streaming, Power BI. Another up-to-30% domain.
• Day 6 — Practice questions. Full mixed sets. Track which of the four domains you're weakest in and re-read that section.
• Day 7 — Final review. Re-run every question you missed, skim the memory aids, and book the exam.

Practice questions are the fastest way to expose the specific distinctions the exam cares about. Run a free DP-900 question set and use your misses to steer your last two days.

Quick FAQ

How much does DP-900 cost?

About $99 USD. Pricing can vary slightly by region.

What's the passing score?

700 on a 1–1000 scale. It's a scaled score — not a straight 70% of questions correct.

How many questions are on the exam?

Roughly 40–60, most commonly around 40–50, in 45 minutes. Mostly multiple-choice and drag-and-drop, with no labs.

Does DP-900 expire?

No. As a Fundamentals certification it has no renewal requirement. Pass once and it's permanent.

Are there prerequisites?

None. No coding, no prior Azure experience.

Is the exam changing in 2026?

There's a minor content refresh effective July 21, 2026. The four domains and their weights stay the same, with slightly more emphasis on Microsoft Fabric and Databricks. The cert is not being retired.

Next step

DP-900 rewards recognition speed — the faster you can map an example to the right concept or service, the more time you have for the tricky ones. The most efficient path there is repetition on real question formats.

Test yourself. Run free DP-900 questions — no card, no email-trap — and turn your weak spots into the last thing you review before test day.