Microsoft DP-900: The Azure Data Fundamentals Exam That Rewards People Who Can Read Data, Not Memorize Service Names

Give most people a spreadsheet of customer orders and they can tell you something useful within a minute: which products sell, which regions lag. Give those same people a list of Azure service names to memorize and they freeze. Here is the good news about DP-900: it is much closer to the first test than the second.

DP-900, the exam behind the Microsoft Certified: Azure Data Fundamentals credential, is routinely filed under "easy Microsoft exam" by people who have never sat it, and routinely underestimated by people who have. Both groups miss the point: this is a reasoning exam dressed up as a fundamentals exam. The questions hand you a situation and ask you to pick the approach that fits. If you can look at data and say what it is and what you would do with it, you are most of the way there.

What the exam actually tests

Microsoft describes DP-900 as a chance to demonstrate knowledge of core data concepts and related Azure data services, aimed at candidates beginning to work with data in the cloud — concepts first, services second. The exam assumes familiarity with relational and non-relational data and with transactional and analytical workloads. It never asks you to configure a service or debug a pipeline.

DP-900 is a concepts exam, not a configuration exam. You will be asked whether a scenario calls for a relational database at all, whether the workload is transactional or analytical, whether the data is structured, semi-structured, or unstructured, and which Azure service matches the shape of the problem. The verbs in the official skills outline give the game away: describe, identify, recognize.

The logistics are manageable. You get 45 minutes of exam time inside a 65-minute appointment, and you need a score of 700 or greater on Microsoft's 1,000-point scaled scale to pass. Microsoft does not publish a fixed question count for DP-900, so any source stating an exact number as fact is guessing. The exam costs $99 in the United States, with the price varying by country or region, and it is delivered proctored through Pearson VUE. There are no prerequisites, and the credential is not a prerequisite for any role-based certification, though Microsoft notes it is useful preparation for the Azure Database Administrator Associate and Azure Data Engineer Associate paths. The skills outline currently in force is dated July 21, 2026, and no retirement has been announced for DP-900.

The four domains, and what each one is really asking

Four domains, and their weights tell you where to spend your time — two together account for roughly half the exam, and even the smallest is worth up to a fifth.

1. Describe core data concepts (25–30%)

The biggest domain alongside analytics, and the one most candidates under-study because it looks easy. It covers how data is represented — structured, semi-structured, and unstructured — file formats and data stores, transactional versus analytical workloads, and the roles on a data team.

This is where the "read data" thesis lives or dies. You need to look at a description of data — a JSON document, a CSV export, a video file, a sensor feed — and classify it correctly. You need to know why a transactional workload (many small, fast writes, like processing orders) behaves differently from an analytical workload (large reads across historical data, like quarterly reporting). Expect scenario questions here, not definitions. If you can explain a concept to a colleague in plain language, you know it well enough.

2. Identify considerations for relational data on Azure (20–25%)

Relational data is the oldest idea in the exam and the one most people think they already know. You need the concepts done properly: what makes data relational, what normalization is and why it is used, common SQL statements, and standard database objects.

Then comes the Azure mapping: the Azure SQL family, including Azure SQL Database, Azure SQL Managed Instance, and SQL Server on Azure Virtual Machines, plus the Azure database services for open-source systems such as PostgreSQL and MySQL. The trap is treating these as interchangeable "SQL in the cloud" options. They are not. Each exists because a different situation calls for it — a fully managed database versus a managed instance with near-complete SQL Server compatibility versus a virtual machine you control yourself. Learn the trade-off, not the product page.

3. Describe considerations for working with non-relational data on Azure (15–20%)

The smallest domain. Not everything fits in rows and columns, and forcing it to fit is the mistake the exam wants you to spot. You need the capabilities of Azure Storage — Blob storage, Azure Files, Table storage — and the capabilities and use cases of Azure Cosmos DB, including its APIs.

The exam's favourite move is to describe a workload — a product catalogue, a session store, a globally distributed app — and ask which store fits. The wrong answers are always plausible if you only know names: Blob storage is not a database, Cosmos DB is not a file share, and Table storage is not a relational table with the SQL removed.

4. Describe an analytics workload on Azure (25–30%)

The other heavyweight domain, and the one that ties the whole exam together. It covers large-scale analytics — ingestion and processing, analytical data stores, Azure Databricks and Microsoft Fabric — plus real-time analytics, batch versus streaming, and data visualization in Power BI, including data models and choosing appropriate visualizations.

This domain rewards people who can trace data through a pipeline in their head: it arrives here, it gets processed there, it lands in an analytical store, someone visualizes it. Note that Microsoft Fabric now sits here alongside Databricks — the July 2026 outline update kept the four domains intact but refreshed the details, a reminder to study from current material.

A practical study approach

Two to four weeks of steady study is enough for most candidates with a general technology background. Start with the official Microsoft Learn path for DP-900, in domain order, but do not just read it — argue with it. After each module, close the page and explain the idea aloud in plain language.

Then build a decision table, one row per Azure service in the outline: what kind of data it holds, what it is good at, what it is bad at, and the scenario that would make you choose it. This single exercise is worth more than a week of flashcards, because it forces you to learn distinctions instead of names — and distinctions are what the exam tests.

Then take practice questions under timed conditions — the real exam gives you 45 minutes, so practice at that pace. For every question you get wrong, write down in one sentence the distinction you missed: "I confused Blob storage with a database." That list of missed distinctions is your study plan for the final days. Working through free DP-900 practice questions before you book the exam will tell you honestly whether you are ready or still memorizing. (Study4Pass DP-900 practice)

Common traps and mistakes

The first trap is over-studying services and under-studying concepts. Candidates walk in able to name every Azure data service and walk out having failed the core concepts domain, worth up to 30 percent. Concepts first, services second.

The second trap is confusing similar services: Azure SQL Database versus SQL Managed Instance, Blob storage versus Data Lake Storage, Cosmos DB versus Table storage, batch versus streaming. The exam is engineered around these pairs, and the difference is always a real trade-off, never trivia.

The third trap is treating DP-900 like AZ-900. The two exams share a format and a passing score, but DP-900 goes deeper on data thinking. The vocabulary is narrower, the scenarios are more specific, and "I know Azure" is not the same as "I understand data."

The fourth trap is stale material. The skills outline was updated on July 21, 2026; older courses may still teach the previous shape of the exam. Check the date on anything you study.

Bottom line

DP-900 is the rare fundamentals exam that measures something employers actually care about: whether you can look at data, understand what kind of problem it presents, and choose the right approach. Study the concepts until you can explain them, learn the services as answers to specific problems rather than names on a list, and practice with scenarios until the distinctions feel obvious. Do that, and the exam stops being a memory test. The people who pass DP-900 comfortably are not the ones who memorized the most service names. They are the ones who learned to read data.