Best Practices
Learn how to get the most effective results from the AI agent by following proven best practices for question writing and analysis workflows.
Writing Effective Questions
Section titled “Writing Effective Questions”✅ Good Questions
- ”What drives customer satisfaction scores?"
- "Show me sales trends over the last year"
- "Which features predict user engagement?"
- "Analyze sentiment in product reviews”
❌ Vague Questions
- ”Tell me about my data"
- "What should I do?"
- "Is this good or bad?"
- "Make a chart”
Getting Better Results
Section titled “Getting Better Results”- Be Specific: Ask about particular relationships or trends
- Provide Context: Mention business goals or concerns
- Ask Follow-ups: Dive deeper into interesting findings
- Review Charts: Check the generated visualizations and data
- Iterate: Refine questions based on initial responses
Question Writing Framework
Section titled “Question Writing Framework”Start with Purpose
Section titled “Start with Purpose”- What business decision are you trying to make?
- What specific insight would be most valuable?
- What action might you take based on the answer?
Be Specific About Variables
Section titled “Be Specific About Variables”- Name specific columns or metrics
- Define time periods clearly
- Specify segments or groups of interest
Provide Business Context
Section titled “Provide Business Context”- Mention your industry or domain
- Explain why this analysis matters
- Share relevant business constraints
Progressive Analysis Strategy
Section titled “Progressive Analysis Strategy”1. Start Broad
Section titled “1. Start Broad”Begin with exploratory questions to understand your data:
- “What are the main patterns in this dataset?”
- “Show me an overview of key metrics”
- “Are there any obvious outliers or anomalies?“
2. Focus on Insights
Section titled “2. Focus on Insights”Drill down into specific areas of interest:
- “What drives the differences in performance?”
- “Which factors correlate with our key outcomes?”
- “How do trends vary across different segments?“
3. Test Hypotheses
Section titled “3. Test Hypotheses”Ask specific questions to validate your assumptions:
- “Does increasing marketing spend improve conversion rates?”
- “Are customers in region X actually more valuable?”
- “Is there a seasonal pattern in our sales data?”
Common Mistakes to Avoid
Section titled “Common Mistakes to Avoid”Vague Requests
Section titled “Vague Requests”- ❌ “Analyze my data”
- ✅ “Show me which customer segments have the highest lifetime value”
Missing Context
Section titled “Missing Context”- ❌ “Is 15% good?”
- ✅ “Is a 15% conversion rate good for our e-commerce industry?”
Single-Shot Analysis
Section titled “Single-Shot Analysis”- ❌ Asking one question and stopping
- ✅ Following up with related questions to build understanding
Maximizing Analysis Quality
Section titled “Maximizing Analysis Quality”Data Preparation
Section titled “Data Preparation”- Clean column names before uploading
- Include relevant time periods
- Ensure data types are correctly formatted
- Remove or flag obvious data quality issues
Question Sequence
Section titled “Question Sequence”- Start with data quality checks
- Move to exploratory analysis
- Focus on specific business questions
- End with actionable insights
Validation
Section titled “Validation”- Cross-check findings with business knowledge
- Review the generated charts and analysis
- Test findings with different time periods or segments
- Consider statistical relevance of patterns
What’s Next?
Section titled “What’s Next?”Chat Management
Learn how to organize multiple analysis conversations and manage your workflow.
Question Types
Explore specific types of analysis questions and when to use each approach.