Key Driver Analysis Skill
Key Driver Analysis: Understand why a metric changed between two time periods. Identifies which dimensions (regions, products, categories, segments) contributed most to revenue growth, sales drops, or other period-over-period changes.
Experimental Feature
This is an experimental feature that is still under active development. Its behavior may change in future releases, or the feature may be removed.
How It Works
The Key Driver Analysis skill compares metric values between two time periods and identifies which dimensions contributed most to the change. When activated, the assistant:
- Identifies the metric and two time periods to compare
- Creates a Key Driver Analysis specification
- Executes the analysis to calculate contributions from each dimension
- Identifies the top contributors (positive and negative)
- With data sharing enabled, provides natural language explanations of what drove the change
The skill helps you understand not just that a metric changed, but specifically which dimensions (regions, products, customer segments, etc.) were responsible for the change, enabling targeted action.
Examples
Understanding revenue growth:
User: "Why did revenue increase from Q1 to Q2?"
Assistant: [Compares periods, identifies top contributing regions/products, explains drivers]Analyzing sales drop:
User: "What caused the sales drop this month compared to last month?"
Assistant: [Runs key driver analysis, shows negative contributors, explains findings]Identifying growth drivers:
User: "Which regions contributed most to the growth?"
Assistant: [Analyzes period comparison, ranks regions by contribution, visualizes results]Other use cases:
- “What caused the sales drop this month?”
- “What products drove the revenue increase?”
- “Compare this quarter to last quarter and show what changed”
- “What dimensions explain the difference between these two periods?”
Limitations
- Requires two distinct time periods for comparison
- Analysis focuses on dimensional contributions, not external factors
- Results interpretation requires data sharing to be enabled (when disabled, provides visualization only)
- Works best when dimensions have sufficient data in both periods