Analysis Process

Watch how the AI agent systematically approaches data analysis with transparent, verifiable methodology. Every step is shown in real-time.

When you ask a question, you’ll see the agent work through these steps:

What happens: The agent analyzes your natural language question to understand:

• The type of analysis you want (correlation, comparison, trend, etc.)
• The business context and goals
• Specific variables or metrics mentioned
• The level of detail needed

What you see:

• “Identifying relevant variables.”

What happens: The agent examines your dataset to find:

• Primary variables relevant to your question
• Supporting variables that might provide context
• Data quality and completeness for each variable
• Relationships between different columns

What you see:

• “Constructing visualizations and transformations over [identified variables].“

What happens: The agent performs the actual calculations:

• Statistical tests and significance calculations
• Correlation analysis or other appropriate methods
• Data aggregation and grouping as needed
• Error checking and validation

What you see:

• “Built X charts. Validating.”
• “Validated X charts. Processing the data transformations.”

What happens: Charts and graphs are generated:

• Automatic selection of appropriate chart types
• Color coding and styling for clarity
• Interactive elements for exploration
• Statistical overlays (trend lines, confidence intervals)

What you see:

• Real-time progress updates during chart generation
• Function headers showing PXL expressions being processed

What happens: Results are interpreted and explained:

• Key findings highlighted with statistical backing
• Business implications and recommendations
• Suggestions for follow-up questions
• Complete methodology documentation

What you see:

• Clear findings with confidence levels
• Actionable recommendations
• Links to detailed methodology
• Suggested next analyses

Every analysis includes:

• Statistical methods used and why they were chosen
• Assumptions made and their validity
• Limitations of the analysis
• Confidence levels and uncertainty measures

All results are verifiable:

• Complete calculation details available on click
• Data filters and transformations documented
• Code equivalent shown for technical users
• Version tracking for analysis reproducibility

The agent provides quality metrics:

• Sample sizes for statistical validity
• Data completeness percentages
• Outlier detection and handling
• Confidence intervals for estimates

For straightforward questions like “What’s our average revenue?”:

• Quick data aggregation
• Basic statistical measures
• Simple visualization
• 5-15 second response time

For multi-faceted questions like “What factors drive customer churn?”:

• Multiple variable analysis
• Advanced statistical testing
• Comprehensive visualizations
• 30-60 second response time

For exploratory questions that build on each other:

• Context preservation across questions
• Progressive depth of analysis
• Connected insights and findings
• Cumulative understanding

When problems are found:

• Missing data: Documented and handled appropriately
• Outliers: Identified and impact assessed
• Data quality: Issues flagged with suggestions
• Insufficient data: Alternative approaches suggested

When constraints are encountered:

• Small sample sizes: Statistical limitations explained
• Correlation vs. causation: Clearly distinguished
• Incomplete data: Impact on conclusions noted
• Methodological constraints: Alternative approaches suggested