DAA-C01 Question 221
Single answerMake predictions based on dataA retail analytics team stores three years of daily sales in Snowflake and wants to forecast next month's sales for each product category directly in SQL. The team prefers a built-in capability that minimizes data movement and does not want to export data to an external ML platform. They also need a solution that can generate future values from historical time-series patterns such as trend and seasonality. Which approach should the data analyst recommend?
- A
Use Snowflake Cortex Forecast to train on the historical sales time series and generate future sales predictions for each product category.
- B
Create a semantic view over the sales table and query it with natural language prompts to produce forecasts.
- C
Use APPROX_PERCENTILE on historical sales by category to estimate next month's sales because percentile functions are designed for predictive time-series modeling.
- D
Use a dynamic table to materialize the last 30 days of sales, because dynamic tables automatically extrapolate future values from recent trends.
Show answer and explanation
Correct answer: A
Explanation
The key requirement is to make predictions about future sales from historical time-series data directly inside Snowflake with minimal operational overhead. The most appropriate recommendation is Snowflake's built-in forecasting capability, which is designed for time-series prediction use cases such as sales forecasting and can account for recurring patterns in historical data. By contrast, semantic views improve data access and governance, percentile functions summarize existing distributions, and dynamic tables maintain transformed datasets; none of these features performs future-value prediction on its own. This aligns with Snowflake best practices of using native in-platform analytics and ML capabilities when the goal is to reduce data movement and operational complexity while generating predictions from data.
- A. Correct.
Correct. Snowflake provides forecasting capabilities intended for predicting future values from historical time-series data directly within Snowflake. This is the best fit for a requirement to forecast next month's sales by category while minimizing data movement and using built-in functionality. Forecasting tools are specifically designed to learn patterns such as trend and seasonality from historical observations and generate future predictions.
- B. Incorrect.
Incorrect. Semantic views help standardize business meaning and improve analytics consumption, and natural language interfaces can assist users in querying data. However, they do not themselves perform time-series forecasting or train predictive models for future-value generation.
- C. Incorrect.
Incorrect. APPROX_PERCENTILE is an aggregate function for estimating percentile values in existing data distributions. It is useful for descriptive statistics, not for predicting future observations in a time series. A candidate might choose this if they confuse statistical summarization with predictive modeling.
- D. Incorrect.
Incorrect. Dynamic tables automate data transformations and incremental refresh of derived datasets. They do not provide predictive modeling or forecast future values. A candidate might select this option because dynamic tables are useful for preparing recent-history features, but they are not the forecasting mechanism itself.