DP-900 Describe an analytics workload on Azure Practice Question

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  Azure Stream Analytics with a reference data input.

  Why it's wrong here

  Azure Stream Analytics is built for continuous, real-time processing of streaming events (for example, from Event Hubs or IoT Hub), not for scheduled batch transformations over files already landed in Blob Storage. A reference data input is only a static lookup used to enrich live streaming records, so it cannot handle schema drift or multi-file aggregation across different websites' clickstream logs. To transform a large set of raw clickstream files on a recurring schedule, you need an orchestration and ETL engine like Data Factory, not a stream-processing job.

• ✓

  Azure Data Factory with a Mapping Data Flow.

  Why this is correct

  Azure Data Factory with a Mapping Data Flow provides visual, code-free data transformation that is compiled into Spark jobs, making it well-suited for batch processing raw clickstream JSON or CSV files from multiple sources. Mapping Data Flows support schema drift by allowing columns to be discovered at runtime and handled through pattern-based mappings, so changes in the clickstream schema do not break the pipeline. ADF can be scheduled with tumbling window or scheduled triggers, enabling recurring, reliable processing of the raw files in Blob Storage.

• ✗

  Azure SQL Database with a stored procedure.

  Why it's wrong here

  Azure SQL Database stored procedures operate on relational tables and cannot natively read or transform JSON documents directly from Blob Storage without extra features like BULK INSERT, OPENROWSET, or PolyBase, all of which require explicit schema definitions. Because clickstream files from multiple websites often have evolving and inconsistent fields, you would need continuous manual ALTER TABLE and script updates, which defeats the purpose of an automated ETL process. Stored procedures also lack trigger-based scheduling tied to file arrival in Blob Storage, so they are not the right vehicle for this scenario.

• ✗

  Azure Logic Apps with a JSON parser.

  Why it's wrong here

  Azure Logic Apps is a low-code integration service designed for orchestrating workflows and connecting SaaS apps, not for heavy, large-scale data transformation of raw clickstream batch files. The built-in JSON parser in a Logic App operates on small message payloads (for example, an individual HTTP request or a single file content) and cannot scale to process thousands of files from multiple websites in a single scheduled run. Logic Apps also lacks schema-drift handling and distributed computing, making it costly and inappropriate as a big data ETL tool.