Note: Data Chat is a private beta feature and may have limitations or changes as it evolves. Reach out to your Funnel contact for feedback and support.
Data Chat works best when you ask clear, specific questions about your data. The way you phrase your questions affects the quality and relevance of the responses you receive. This article explains best practices for writing effective prompts and refining your questions to get better results.
Learn more: For step-by-step instructions on using Data Chat, see Getting started with Data Chat.
Writing effective prompts
Clear and specific prompts help Data Chat understand what you need and generate more accurate results. Follow these guidelines when asking questions:
Be specific about what you want
Include details about metrics, dimensions, time periods, and filters in your questions.
Good examples:
Show me spend by channel for the last 30 daysCompare ROAS across my Facebook Ads campaigns from January to MarchWhat's my total conversions for social media channels this month?
Less effective examples:
Too vague:
Show me my dataMissing context:
What happened?Not specific enough:
Give me a report
Include time periods
Specify the date range you want to analyze.
Good examples:
Show me impressions for the last 7 daysCompare cost per click between Q1 and Q2 this yearWhat's my conversion trend for the past 90 days?
Less effective examples:
No time period:
Show me impressionsMissing date context:
What's my performance?
Name specific metrics and dimensions
Use the exact names of metrics and dimensions from your workspace when possible.
Good examples:
Show me ROAS by campaign for Google AdsCompare clicks and conversions across my ad groupsWhat's my cost per acquisition for Facebook Ads?
Less effective examples:
Not specific:
Show me some numbersToo general:
Give me performance data
Ask one question at a time
Focus on a single analysis per question to get clearer results.
Good examples:
Show me spend by channelWhat's my top-performing campaign by conversions?
Less effective examples:
Too many questions at once:
Show me spend by channel and also conversions by campaign and ROAS trends
Filter out conditions you don't want
Specify what to exclude or limit your analysis to specific platforms, campaigns, or other dimensions.
Good examples:
Show me performance for Meta platforms onlyCompare ROAS for Campaign A and Campaign B onlyWhat's my spend for Facebook Ads and Instagram Ads, excluding other channels?Show me conversions for social media channels, not search
Less effective examples:
Doesn't specify what to include or exclude:
Show me my performanceToo broad, includes everything:
Give me campaign data
Prompting examples by use case
Use these examples as starting points for common analysis tasks:
Performance overview
Give me a performance overview of my Facebook Ads campaigns for the last 30 daysShow me total spend, clicks, and conversions by channel this monthWhat's my overall marketing performance compared to last month?
Channel comparison
Compare ROAS across all my advertising channels for Q1Which channel has the highest cost per acquisition?Show me spend and conversions side-by-side for social media and search
Trend analysis
What's my impression trend for the last 7 days?Show me how my conversion rate changed over the past monthCompare my spend trends week-over-week
Campaign analysis
Show me my top 10 campaigns by conversionsWhich campaigns have the lowest ROAS?Compare performance metrics for all active campaigns
Custom questions
Show me cost per click for campaigns with more than 1000 clicksWhat's my conversion rate for mobile traffic only?Compare performance for campaigns created in the last 30 days
Refining your prompts
If Data Chat doesn't return the results you expected, refine your question using these strategies:
Add more context
Include additional details about what you're looking for.
Original prompt: Show me my campaigns
Refined prompt: Show me my Facebook Ads campaigns with spend greater than $1000 for the last 30 days
Specify the format you want
Tell Data Chat what type of visualization or breakdown you prefer.
Original prompt: Show me spend by channel
Refined prompt: Show me spend by channel as a bar chart with a table breakdown
Clarify metrics and dimensions
Use exact metric and dimension names from your workspace.
Original prompt: Show me performance
Refined prompt: Show me cost, clicks, conversions, and ROAS by campaign for Google Ads
Break down complex questions
Split multi-part questions into separate, focused queries.
Original prompt: Show me spend by channel and conversions by campaign and also ROAS trends
Refined prompts:
Show me spend by channel for the last 30 daysShow me conversions by campaign for the same periodShow me ROAS trends over the last 30 days
Use follow-up questions
Ask clarifying questions in the same chat to refine results.
Example conversation:
First question:
Show me spend by channelFollow-up:
Which channel has the highest spend?Follow-up:
Show me conversions for that channel only
Common prompting patterns
These patterns work well for different types of analysis:
Comparison requests
Pattern: Compare [metric] across [dimension] for [time period]
Examples:
Compare ROAS across my advertising channels for Q1Compare cost per click between Facebook Ads and Google Ads this month
Ranking requests
Pattern: Show me [top/bottom] [number] [items] by [metric]
Examples:
Show me my top 10 campaigns by conversionsWhat are my bottom 5 ad groups by ROAS?
Trend requests
Pattern: Show me [metric] trend for [time period]
Examples:
Show me my conversion rate trend for the last 30 daysWhat's my spend trend week-over-week?
Filtered analysis
Pattern: Show me [metric] for [dimension] where [condition]
Examples:
Show me conversions for campaigns with spend greater than $5000What's my ROAS for social media channels only?
Tips for better results
Verify the responses: Data Chat can make mistakes sometimes. If you don't get the correct response, give more context and details in your prompts.
Start broad, then narrow: Begin with a general question, then use follow-ups to dive deeper into specific areas.
Use workspace terminology: Data Chat understands your workspace context, including custom fields and naming conventions.
Set up workspace instructions: Use workspace instructions to customize how Data Chat responds across your workspace. Define business context, terminology, and preferred metrics to get more consistent, relevant responses.
Reference previous results: In follow-up questions, you can reference previous answers to build on your analysis.
Provide feedback: Use the thumbs-up or thumbs-down icons to rate responses, which helps Data Chat learn your preferences.
Understand how responses are created: Click Thought for a few seconds below your prompt to see how Data Chat frames and creates its response.
Open in Data Explorer: For complex analysis, click Open in data explorer to modify filters and add more dimensions.
What to avoid
Overly complex questions: Break down multi-part questions into separate queries.
Vague requests: Be specific about metrics, dimensions, and time periods.
Ambiguous language: Use clear, direct language instead of jargon or abbreviations unless they're standard in your workspace.
Asking for actions: Data Chat is read-only and cannot make changes to settings or data.