Overview
Domain 4 covers the analytics side of data on Azure through three functional groups. The first is large-scale analytics: considerations for ingesting and processing data, options for analytical stores such as data warehouses, data lakes and lakehouses, and the Microsoft cloud services for analytics, which the outline names as Azure Databricks and Microsoft Fabric.
The second group is real-time analytics: how batch data differs from streaming data and which Microsoft services handle streams, including Azure Event Hubs, Azure Stream Analytics and Real-Time Intelligence in Fabric. The third group is Power BI: its capabilities across Desktop and the service, the features of its data models such as tables, relationships, measures and hierarchies, and choosing the right visual for a question. Together with Domain 1, it is the joint largest part of the exam.
How it's tested
Questions describe a reporting, ingestion or monitoring requirement and ask which store, service, processing style or visual fits. Expect wording about latency, data volume, open file formats, SQL versus Spark skills, and the business question a chart should answer. Product names in options are usually current Fabric, Databricks and Power BI terms.
-
Choose the analytical store
Relational SQL reporting points to a data warehouse, raw files in many formats to a data lake, and both together on open Delta tables to a lakehouse.
-
Batch or stream
Results needed within seconds from continuous events indicate streaming services, while scheduled processing of complete datasets indicates batch pipelines.
-
Match the visual
Trends over time suit line charts, comparisons across categories suit bar or column charts, and single key figures suit cards.
A logistics firm wants a dashboard showing truck locations and alerts within seconds of each GPS event. A nightly pipeline into a warehouse would leave data hours old, so batch processing fails the latency need. Events should flow through an ingestion service such as Azure Event Hubs or a Fabric eventstream, into Real-Time Intelligence or Stream Analytics for continuous queries, then on to a live dashboard. The deciding clue is seconds of latency on unbounded data.
Functional groups 3
Official functional groups from the Microsoft outline, in outline order.
4.1 Describe common elements of large-scale analyticsIngestion and processing, analytical data stores, and Azure Databricks and Microsoft Fabric.
- What it means
- This group covers considerations for data ingestion and processing, options for analytical data stores, and Microsoft cloud services for large-scale analytics, including Azure Databricks and Microsoft Fabric.
- What to understand
- Ingestion copies data from operational sources into an analytical store, often with pipelines in Fabric Data Factory or Azure Data Factory, using ETL or ELT. Processing cleans, joins and aggregates data, typically with SQL or Apache Spark. A data warehouse stores modeled relational data, often as a star schema of fact and dimension tables, for SQL reporting. A data lake holds raw files in many formats cheaply. A lakehouse combines both, storing tables in open Delta Parquet format on lake storage while supporting SQL queries. Azure Databricks is a Spark-based platform for data engineering, analytics and machine learning. Microsoft Fabric is a SaaS analytics platform with OneLake as a single data lake and workloads for Data Factory, lakehouses, warehouses, Real-Time Intelligence and Power BI.
- How it shows up
- Expect a requirement to classify as warehouse, lake or lakehouse, or to identify Fabric or Databricks as the right platform.
- Common confusion
- Thinking a lakehouse is just a data lake; it adds table structure, transactions and SQL access over open files.
4.2 Describe considerations for real-time data analyticsBatch versus streaming data and Microsoft real-time analytics services.
- What it means
- This group covers describing the difference between batch and streaming data and identifying Microsoft cloud services for real-time analytics.
- What to understand
- Batch processing collects data over a period and processes the complete set on a schedule, giving high throughput but results that are minutes or hours old. Stream processing handles each event or small window as it arrives from an unbounded source, giving results in seconds or less. Streaming queries often aggregate over time windows such as tumbling or hopping windows. Azure Event Hubs ingests high volumes of events. Azure Stream Analytics runs SQL-like queries over streams. In Microsoft Fabric, Real-Time Intelligence provides eventstreams for capture and routing, eventhouses with KQL databases for storage and querying, real-time dashboards and Activator for triggering actions. Azure Databricks supports streaming through Spark Structured Streaming.
- How it shows up
- Expect a latency and source description, and you choose batch or streaming, or the service that ingests or analyzes the stream.
- Common confusion
- Choosing streaming whenever data is large; volume alone does not require streaming if results can wait.
4.3 Describe data visualization in Microsoft Power BIPower BI capabilities, data model features and appropriate visuals.
- What it means
- This group covers identifying the capabilities of Power BI, describing features of data models in Power BI, and identifying appropriate visualizations for data.
- What to understand
- Power BI Desktop is used to connect to sources, shape data with Power Query, build a model and design reports. The Power BI service, part of Microsoft Fabric, publishes reports, builds dashboards from pinned visuals, manages workspaces, sharing and scheduled refresh, and Power BI Mobile shows content on phones. A semantic model contains tables, relationships with cardinality and filter direction, calculated columns, measures written in DAX, and hierarchies such as year, quarter and month for drill-down. Star schemas with a fact table and related dimension tables work best. Visual choice follows the question: line charts for trends, bar or column charts for comparisons, cards for single values, maps for geography, matrices for detail and scatter charts for correlations.
- How it shows up
- Expect a reporting need and a choice of model feature, Power BI component or visual type.
- Common confusion
- Confusing a report, which can have many pages, with a dashboard, a single page of pinned tiles in the service.
High-yield concepts
The terms, rules and formulas this domain turns into questions most often.
-
Data warehouse
Relational store of modeled data for analytics.
- Remember
- Star schemas and SQL queries.
- Tested as
- Analytical store questions.
- Trap
- Using it for raw unstructured files.
-
Data lake
Low-cost store for raw files of any format.
- Remember
- Schema applied when data is read.
- Tested as
- Analytical store questions.
- Trap
- Expecting enforced table schemas.
-
Lakehouse
Lake storage with table structure and SQL access.
- Remember
- Open Delta Parquet tables support transactions.
- Tested as
- Fabric and Databricks questions.
- Trap
- Treating it as only a folder of files.
-
Microsoft Fabric
SaaS analytics platform built on OneLake.
- Remember
- Data Factory, lakehouse, warehouse, Real-Time Intelligence, Power BI.
- Tested as
- Platform questions.
- Trap
- Thinking each workload needs its own storage copy.
-
OneLake
Single logical data lake for a Fabric tenant.
- Remember
- All Fabric items store data there in open formats.
- Tested as
- Fabric storage questions.
- Trap
- Confusing it with a separate database.
-
Azure Databricks
Managed Apache Spark analytics platform.
- Remember
- Notebooks for engineering, analytics and machine learning.
- Tested as
- Platform questions.
- Trap
- Assuming it is a reporting tool.
-
Star schema
Fact table linked to dimension tables.
- Remember
- Facts hold measures, dimensions hold descriptive attributes.
- Tested as
- Modeling questions.
- Trap
- Putting descriptive text in fact tables.
-
Batch processing
Processing a complete dataset on a schedule.
- Remember
- High throughput with delayed results.
- Tested as
- Processing style questions.
- Trap
- Choosing it for second-level latency needs.
-
Stream processing
Processing unbounded events as they arrive.
- Remember
- Low latency, often over time windows.
- Tested as
- Real-time questions.
- Trap
- Using it when daily reports suffice.
-
Azure Event Hubs
Big data event ingestion service.
- Remember
- Receives millions of events for downstream processing.
- Tested as
- Real-time ingestion questions.
- Trap
- Expecting it to run analytical queries.
-
Real-Time Intelligence
Fabric workload for streaming data.
- Remember
- Eventstreams, eventhouses with KQL, real-time dashboards, Activator.
- Tested as
- Real-time service questions.
- Trap
- Confusing an eventhouse with a warehouse.
-
Measure
DAX calculation evaluated at query time.
- Remember
- Responds to filters and slicers in visuals.
- Tested as
- Power BI model questions.
- Trap
- Using a calculated column for dynamic totals.
-
Hierarchy
Ordered levels such as year, quarter, month.
- Remember
- Enables drill-down in visuals.
- Tested as
- Power BI model questions.
- Trap
- Confusing it with a relationship.
-
Power BI dashboard
Single-page canvas of pinned tiles in the service.
- Remember
- Built from visuals on one or more reports.
- Tested as
- Power BI capability questions.
- Trap
- Thinking dashboards are built in Desktop.
Side by side
Concepts candidates mix up, compared on the point that decides the answer.
Analytical data stores
| Concept | Meaning | Key difference | Common trap |
|---|---|---|---|
| Data warehouse | Relational tables for modeled data | Schema on write, SQL-first reporting | Loading raw images or logs into it |
| Data lake | Files in open or raw formats | Schema on read, cheap and flexible | Expecting transactions on plain files |
| Lakehouse | Delta tables on lake storage | Lake flexibility with table and SQL features | Thinking it requires a separate warehouse copy |
Batch versus streaming
| Concept | Meaning | Key difference | Common trap |
|---|---|---|---|
| Batch | Bounded data processed on a schedule | Results arrive after the batch completes | Using it for live monitoring |
| Streaming | Unbounded events processed continuously | Results within seconds, windowed aggregates | Picking it for monthly reports |
Power BI building blocks
| Concept | Meaning | Key difference | Common trap |
|---|---|---|---|
| Semantic model | Tables, relationships, measures and hierarchies | Shared data layer behind reports | Confusing it with a report |
| Report | Multi-page interactive visuals | Built in Desktop or the service from one model | Calling it a dashboard |
| Dashboard | Single page of pinned tiles | Created only in the Power BI service | Expecting full report interactivity |
Choosing a visual
| Concept | Meaning | Key difference | Common trap |
|---|---|---|---|
| Line chart | Values plotted over time | Shows trends and seasonality | Using it for unordered categories |
| Bar or column chart | Lengths compared across categories | Shows ranking and comparison | Using it for a single KPI |
| Card | One prominent number | Shows a key figure such as total sales | Packing many values into cards |
| Map | Values placed by location | Shows geographic patterns | Using it without location fields |
Exam traps
Each trap is an answer that looks right. Learn the clue that should change your mind.
-
Warehouse for raw files
- What candidates pick
- Load raw JSON, images and logs straight into a data warehouse.
- Why it looks right
- Warehouses are for analytics.
- Why it's wrong
- Warehouses hold modeled relational tables; raw multi-format files belong in a data lake or lakehouse.
Clue to watch for Data in many raw formats awaiting processing.
-
Streaming because of volume
- What candidates pick
- Use streaming to process a 5 TB nightly export.
- Why it looks right
- Large data sounds like it needs real time.
- Why it's wrong
- Bounded data with no latency requirement is processed most efficiently in batch.
Clue to watch for A nightly or scheduled delivery.
-
Event Hubs runs queries
- What candidates pick
- Use Event Hubs alone to calculate rolling averages.
- Why it looks right
- It handles streaming data.
- Why it's wrong
- Event Hubs ingests events; Stream Analytics, Spark or an eventhouse performs the analysis.
Clue to watch for The requirement is to compute or query, not only to receive.
-
Dashboard in Desktop
- What candidates pick
- Build the dashboard in Power BI Desktop.
- Why it looks right
- Desktop is where reports are built.
- Why it's wrong
- Dashboards are created in the Power BI service by pinning visuals from published reports.
Clue to watch for The word dashboard with pinned tiles.
-
Calculated column for totals
- What candidates pick
- Create a calculated column to show sales that change with slicers.
- Why it looks right
- Columns feel simpler.
- Why it's wrong
- Calculated columns are computed at refresh per row; measures recalculate for each filter context.
Clue to watch for A value that must react to user selections.
-
Pie chart for trends
- What candidates pick
- Show monthly revenue over two years with a pie chart.
- Why it looks right
- Pie charts are familiar.
- Why it's wrong
- Time trends need an ordered axis, so a line chart communicates them clearly.
Clue to watch for The phrase over time or by month.
-
Databricks as BI
- What candidates pick
- Use Azure Databricks to publish interactive business reports to managers.
- Why it looks right
- It handles analytics.
- Why it's wrong
- Databricks processes and analyzes data; Power BI is the service for business reports and dashboards.
Clue to watch for End users need interactive visual reports.
Answer framework
Work through these steps for analytics store, processing and Power BI questions in this domain.
- Determine whether data arrives as bounded sets on a schedule or as a continuous stream, and how quickly results are needed.
- For storage, pick a warehouse for modeled SQL tables, a data lake for raw files, or a lakehouse for both on Delta tables.
- For platform, recognize Microsoft Fabric as the unified SaaS option on OneLake and Azure Databricks as the Spark-based platform.
- For real-time needs, map ingestion to Event Hubs or eventstreams and analysis to Stream Analytics or Real-Time Intelligence eventhouses.
- For Power BI, decide whether the need is a model feature, a report or dashboard, or a visual, then match the question asked.
Practice question
Want more like this? The DP-900 practice test has questions for every domain.
Checklist
0 of 13
Tick each one you can do without notes. Your ticks stay in this browser.
Study strategy
Give this domain the most study time along with Domain 1, and use current product names from the start. Read the Microsoft Learn modules on large-scale analytics, real-time analytics and Power BI, and draw one diagram that shows sources flowing through ingestion into a lake, lakehouse or warehouse, then into a semantic model and reports. Label where Fabric Data Factory, OneLake, Azure Databricks and Power BI sit.
If possible, start a Fabric trial, create a lakehouse, load a CSV, view it as a Delta table and query it with the SQL analytics endpoint. Then explore Real-Time Intelligence with the sample eventstream to see streaming data land in an eventhouse. For Power BI, install Desktop, import a small sales dataset, create relationships, write one measure, build a hierarchy, and try several visuals for the same data to see which answers which question. Finish with scenario drills on batch versus streaming and visual choice, since these appear in many forms.
FAQ
How much of DP-900 is Domain 4, Describe an analytics workload?
Microsoft weights Describe an analytics workload at 25–30% of DP-900, tied with core data concepts as the largest domain, across three functional groups.
Is Azure Synapse Analytics still tested on DP-900?
The current outline names Azure Databricks and Microsoft Fabric as the large-scale analytics services. Focus your study there rather than on Synapse-specific features.
How does a lakehouse differ from a plain data lake?
A data lake stores raw files. A lakehouse adds table structure in Delta format, transactions and SQL access on top of lake storage.
Which services handle real-time data?
Azure Event Hubs ingests events, Azure Stream Analytics queries streams, and Fabric Real-Time Intelligence provides eventstreams, eventhouses and real-time dashboards.
Do I need to write DAX for DP-900?
No detailed DAX is required. You should know what a measure does and when it is preferred over a calculated column.
Where are Power BI dashboards created?
In the Power BI service, by pinning visuals from published reports. Reports themselves are usually authored in Power BI Desktop.
Other Data & Analytics certifications
Where the same subject shows up in other exams. Cards marked "Same subject" open the matching domain.
Test your Domain 4 knowledge
You have covered ingestion, warehouses and lakehouses, Fabric and Databricks, streaming services and Power BI. Check whether you can pick the right option under exam conditions.
Keep going
- Training Training Built on your Weak areas
- Practice test Questions by domain, with explanations
- Mock exam Full length, timed like the real exam
- Readiness test Find your weakest domain in minutes
- Cheat sheet Key terms and formulas on one page
- Ask the DP-900 AI tutor Get any concept explained your way