COF-C03 Question 132
Single answerCortex AnalystA retail analytics team wants business users to ask questions such as "What were online sales in the Northeast last quarter?" using Snowflake Cortex Analyst. The semantic model has already been created, but early testing shows the assistant sometimes selects the wrong date field and occasionally treats "online sales" as total sales. Which action should the data team take to improve the quality of SQL generated by Cortex Analyst while keeping the solution maintainable?
- A
Refine the semantic model by adding clear business-friendly metric and dimension definitions, synonyms, and disambiguating metadata for the relevant date and sales fields
- B
Increase the size of the virtual warehouse used by the application so Cortex Analyst has more compute available for natural language understanding
- C
Replace the semantic model with secure views only, because Cortex Analyst relies entirely on view names and column names rather than semantic definitions
- D
Move the sales data into an unstructured stage and use document AI features first, because Cortex Analyst works best when source data is stored as files
Show answer and explanation
Correct answer: A
Explanation
The best answer is to improve the semantic model. Cortex Analyst is built to answer natural-language questions over structured enterprise data by using semantic metadata that defines metrics, dimensions, relationships, and business terminology. When generated SQL is inaccurate because terms are ambiguous, the recommended practice is to refine the semantic model with clearer definitions and synonyms so the service can map user intent correctly. Increasing warehouse size helps with query performance but not language-to-SQL accuracy. Likewise, relying only on views or converting data into unstructured files misunderstands Cortex Analyst's role. For SnowPro Core, candidates should recognize that semantic modeling is the primary control point for improving Cortex Analyst accuracy and maintainability.
- A. Correct.
Correct. Cortex Analyst is designed to use a semantic model to map natural-language business questions to the right tables, dimensions, and metrics. If users say "online sales" or "last quarter," the best way to improve generated SQL is to enrich the semantic model with precise definitions, aliases/synonyms, and metadata that reduce ambiguity. This is the maintainable, intended approach for improving answer quality in realistic deployments.
- B. Incorrect.
Incorrect. Warehouse size affects query execution performance, not the semantic interpretation quality of Cortex Analyst. If the problem is that the assistant chooses the wrong field or misunderstands a business term, adding compute does not address the root cause. This distractor reflects the common misconception that all AI quality issues are solved by scaling compute.
- C. Incorrect.
Incorrect. Cortex Analyst does not depend only on object names. Its purpose is to use a semantic model to understand business meaning beyond raw schema names. Secure views can still be useful for governance or simplification, but replacing the semantic model with views does not solve the core ambiguity problem and removes the mechanism specifically intended to guide Analyst.
- D. Incorrect.
Incorrect. Cortex Analyst is intended for structured data analysis using a semantic layer. Moving relational sales data to an unstructured stage and invoking document-oriented capabilities would add unnecessary complexity and would not improve SQL generation for structured analytics questions. This option confuses Cortex Analyst with other AI capabilities in Snowflake.