DAA-C01 Question 208
Single answer3.3 Perform diagnostic analyses.A retail analytics team uses Snowflake to investigate a sudden 12% drop in weekly online conversions. They have a fact table with order events, a customer dimension, and web session data. Initial KPI dashboards confirm the drop started three days ago, but the business team cannot tell whether the issue is tied to a marketing channel, device type, or a specific customer segment. The analyst needs to perform diagnostic analysis in Snowflake to identify the most likely drivers of the decline before deeper remediation work begins. Which approach should the analyst take?
- A
Drill down into the affected KPI by relevant dimensions such as acquisition channel, device type, geography, and customer segment; compare current versus prior periods; and quantify contribution to the decline using segmented aggregates.
- B
Replace all NULL values in the source tables with zeroes and rerun the dashboard so that missing data does not affect the conversion-rate calculation.
- C
Build a machine learning model to predict conversions for the next four weeks and use the forecast error to explain the recent decline.
- D
Increase the size of the virtual warehouse and rerun the SQL queries so the analyst can determine whether compute performance caused the drop in conversions.
Show answer and explanation
Correct answer: A
Explanation
This question tests the candidate's ability to distinguish diagnostic analysis from adjacent analytics tasks. In a diagnostic workflow, the analyst begins with a known outcome change and investigates contributing factors through segmentation, drill-down, comparison across dimensions, and variance decomposition. In Snowflake, this is typically done by querying fact tables with joins to dimensions and comparing current versus baseline periods using aggregates such as conversion counts, rates, and percentage-point deltas. This aligns with standard analytics best practices: descriptive analysis answers what happened, diagnostic analysis answers why it happened, predictive analysis estimates what will happen, and prescriptive analysis recommends actions. Snowflake supports diagnostic analysis through scalable SQL, secure access to integrated data sets, and analytic functions for comparing time periods and segment performance. The best answer is the one that directly isolates likely causal drivers from the available business dimensions rather than changing data values, forecasting future outcomes, or tuning compute resources.
- A. Correct.
Correct. Diagnostic analysis focuses on understanding why a change occurred by breaking the metric into meaningful dimensions, comparing segments across time periods, and identifying which subpopulations contributed most to the change. In Snowflake, this commonly involves SQL-based slice-and-dice analysis, drill-downs, variance analysis, and joins across fact and dimension tables. This approach is practical and aligned to the scenario because the team already knows what changed and now needs to isolate likely drivers such as channel, device, or customer segment.
- B. Incorrect.
Incorrect. Replacing NULLs with zeroes is a data-cleaning action, not a diagnostic method for identifying the source of the KPI decline. It may also distort business metrics if NULL means unknown, not zero. A candidate might choose this option because data quality can affect analysis, but indiscriminately imputing zeroes is not a best practice and does not directly answer which segment caused the drop.
- C. Incorrect.
Incorrect. Predictive modeling addresses what is likely to happen next, not why the current drop happened. Forecast error may show that performance deviated from expectation, but it does not isolate root contributors across dimensions. This is a common misconception: using advanced analytics when a structured diagnostic drill-down is the appropriate first step.
- D. Incorrect.
Incorrect. Increasing warehouse size may improve query runtime, but it does not explain a business conversion decline. This option confuses system-performance troubleshooting with business diagnostic analysis. Unless there is evidence that dashboards are incomplete or failing due to compute constraints, scaling the warehouse does not help identify the business drivers behind the KPI movement.