Orders & Revenue
Questions about sales, order volume, and revenue metrics.Customers
Questions about customer acquisition, retention, and behavior.Marketing & Ads
Questions about advertising performance and marketing spend.Raw Data Pulls
Questions where you want the underlying rows or specific fields returned as a CSV.- The time range
- The fields or entity you want
- Any filters, such as product, campaign, store, or channel
Use raw data pull wording when you want the data itself. If you want interpretation, ask an analysis question like “What are the trends in orders by UTM source?”
Email & SMS
Questions about outbound messaging performance.Products
Questions about product performance and mix.Cohorts & LTV
Questions about customer lifetime value and cohort behavior.Definitions & Schema
Questions about what metrics mean, where data lives, and how SourceMedium works. These are answered using our documentation — no SQL required.Diagnostics
Health checks for your data and tracking.Learn about Diagnostics
Understand Data Health and Attribution Health checks in detail.
Thread follow-up examples
After you run analysis, you can continue in the same thread with prompts like:- In DM, ask these naturally.
- In channels, mention the bot with
@SourceMedium.
Thread continuity and follow-ups
Learn how follow-up context works and when to start a new thread.
Tips for Better Results
Specify time ranges explicitly
Specify time ranges explicitly
Instead of “recently” or “lately,” use specific ranges like “last 7 days,” “in January,” or “Q4 2024.”
Name the metric you want
Name the metric you want
“Revenue” is clearer than “sales.” “Orders” is clearer than “transactions.” Use the terminology from your dashboards.
Ask one thing at a time
Ask one thing at a time
Complex multi-part questions work better as separate queries. Ask “What was our revenue last month?” then “How does that compare to the previous month?”
Use pull/export language for raw data
Use pull/export language for raw data
“Pull order ID, date, UTM source, and revenue for last week” tells the AI Analyst to return rows and attach a CSV instead of writing an interpretive analysis.
Use 'top N' for rankings
Use 'top N' for rankings
“Top 10 products by revenue” is specific. “Best products” is ambiguous.