Prompting Data Conversations#
This guide explains how to write effective prompts to get insights from your data with Data Conversations.
Prompting best practices#
To get the most out of Data Conversations, follow these guidelines when creating your prompts.
- Be clear about the information you need
Be specific in your questions. Include as many details as possible about the information you’re interested in.
- DO
Show the top 5 campaigns by conversions in January 2025.
- DON’T
Show campaign performance.
- Add timeframes and filters
Context helps obtain the most relevant results. Include specific time periods, accounts, or dimensions in your prompt:
What were the total spendings by country for Q1 2025?
Compare CTR across Facebook and Instagram campaigns in March 2025.
- Use exact column names where possible
For better accuracy, use the exact column names of your target fields in the Data Dictionary, for example,
channel
,campaign_name
, ortotal_impressions
. If unsure, describe the metric or dimension clearly. The system will attempt to interpret terms or calculated metrics it doesn’t immediately recognize and may ask you to clarify. If uncertain about exact names, consult your data team or existing reporting dashboards.- Start a new conversation for new topics
When you’re ready to explore a new topic or analysis, it’s best to start a new conversation. This ensures the previous context, like filters or past questions, doesn’t unintentionally affect your new request.
- Keep your question focused
Stick to one question at a time. Break complex requests into smaller prompts.
For example, first, ask for performance by channel, then ask for trends over time.
- Use follow-up prompts
You’ll see suggested follow-up questions to help you dig deeper into your data and generate more insights. You can also use your own follow-ups:
What about last month?
Break it down by device type.
Prompt examples#
Here are some best practice Data Conversations prompts for typical use cases. You can adjust these examples as needed based on your data and the questions you have.
- Getting an overview
Can you provide summary statistics of the data I have available?
What are the impressions coming from the Facebook Ads?
Is there any data for client, brand, or region [insert name] for [timeframe, e.g., Q4 2024]?
- Basic Performance Analysis
Show me the total ad spend for April 2025.
How many conversions did we get last month from Campaign [insert name]?
What's the average CTR per campaign in Q1?
What are the top 10 keywords by conversion rate this year?
- Segment Deep-dive
Break down spend by channel in January 2025.
Show the top 10 campaigns by conversions for Facebook.
Compare leads by country for Q4 2024.
Show CTR by device type for Facebook campaigns in Q1.
Compare conversion rates across channels for the past 30 days.
- Uncovering Trends
Show a line chart of daily clicks in February.
Show the trend of conversions by week over the last 3 months.
How did cost-per-click change from January to March per campaign?
Show the trend of the website traffic over the past 6 months.
What's the trend in Cost per Acquisition for Google Ads campaigns this month?
- Visual Insights
Show a bar chart of top campaigns by ROAS.
Visualize bounce rate by device type.
Create a chart for weekly spend for Google Ads in 2024.
- E-Commerce Insights
Which products were the top sellers on Shopify over the last month?
Show me the sales trend for the past 12 months.
Compare sales performance across different regions.
- Behavioral Insights (if CRM data is available)
What's the customer lifetime value by acquisition channel?
Show me the customer retention rate by segment.
What's our customer churn rate trend over the past year?
As you use Data Conversations more, you’ll develop a deeper understanding of your data and uncover valuable insights. Don’t hesitate to experiment with different questions - the system is designed to help you explore and learn.