DAA-C01 Question 16
Single answer1.2 Perform data discovery to identify what is needed from the available datasets.A retail analytics team is asked to build a weekly executive dashboard showing net sales by region, product category, and promotion. The data engineer says the necessary information likely exists across several Snowflake tables in different schemas, but the analyst is not sure which tables contain the required attributes or whether key columns are populated consistently enough to support the dashboard. Before designing joins or writing transformation logic, what is the BEST first step to perform data discovery in Snowflake?
- A
Use Snowsight or SQL to inspect INFORMATION_SCHEMA and table metadata, then profile candidate tables with exploratory queries for column availability, null rates, distinct values, and join key coverage.
- B
Create a materialized view that joins all suspected source tables, then use dashboard errors to determine which columns are missing or low quality.
- C
Immediately train a forecasting model on the sales table to identify whether region and promotion columns are sufficiently predictive for executive reporting.
- D
Clone the production database and enable search optimization on every table so the analyst can discover the required data more quickly.
Show answer and explanation
Correct answer: A
Explanation
The best first step in this scenario is structured data discovery: identify candidate tables, inspect metadata, and profile the data before designing joins or transformations. In Snowflake, analysts commonly use INFORMATION_SCHEMA views and Snowsight to discover schemas, tables, columns, and object definitions. After identifying likely sources, exploratory SQL should validate practical usability: whether required columns exist, whether keys are populated, whether dimensions have expected cardinality, and whether values align across datasets for joining. This approach follows analytics best practices because it reduces rework and ensures dashboard requirements are matched to actual available data. Snowflake documentation on INFORMATION_SCHEMA, DESCRIBE commands, and exploratory querying supports this metadata-first workflow for understanding available datasets.
- A. Correct.
Correct. For data discovery, the analyst should first identify candidate datasets and understand their structure and quality before building semantic logic. In Snowflake, this typically means reviewing metadata from INFORMATION_SCHEMA views (such as TABLES and COLUMNS) or Snowsight object details, then running profiling queries to validate whether needed columns exist and whether they are usable. Checking null rates, distinct values, ranges, and join key coverage is essential to determine if the available datasets can support the required dashboard dimensions and metrics.
- B. Incorrect.
Incorrect. Building a materialized view before validating source suitability is premature and inefficient. Materialized views are intended to optimize repeated query patterns, not to discover whether the required attributes exist or whether source keys are reliable. Using downstream failures to infer missing columns is poor analytical practice and increases development time.
- C. Incorrect.
Incorrect. Predictive modeling does not address the core discovery question. The analyst first needs to confirm what data exists, how it is structured, and whether it is complete enough for reporting. Whether a column is predictive is irrelevant to determining whether it can support a weekly executive dashboard grouped by required business dimensions.
- D. Incorrect.
Incorrect. Zero-copy cloning can be useful for safe experimentation, and search optimization can improve some query access patterns, but neither is the best first step for understanding dataset suitability. Enabling search optimization broadly would also add cost and does not solve the fundamental need to inspect metadata and profile data quality.