Analyze your data

Before starting, connect your datasource and complete setup for the app, CLI or MCP. You need permission to query the data used in your analysis.

1. Pick a familiar question

Choose something specific that you can compare with an existing report or query:

• What was revenue last month?
• How many customers signed up last week?
• Which products had the most orders last quarter?

Start with one metric and a defined period. Keep the trusted result and its definition nearby, including any filters and when the data was last refreshed.

2. Give Querio the business meaning

Check that Querio has the context needed to answer your question: what the relevant tables contain, how they relate, and how your team defines the metric.

For example, your revenue definition might be:

Revenue is the sum of paid, non-refunded orders, assigned to the payment date in UTC.

Add reusable definitions as context so your team doesn't have to explain them in every conversation. Put requirements specific to this analysis directly in your question.

You don't need to document your entire business before starting. Begin with the tables and rules needed for this question.

3. Ask the question

In the app, start New Explore, select your datasource and send your question. Through CLI or MCP, give the task to your connected agent and identify the datasource it should use.

For example:

Calculate September 2026 revenue in USD from orders.

Sum amount where status = 'paid' and refunded = false.
Use paid_at in UTC, from September 1 inclusive to
October 1 exclusive. Each row represents one order.

Show the query, total revenue, and a sample of the included
orders in the notebook. Run the analysis.

Replace the table, fields and definition with those used by your team.

4. Check the answer

Read the answer alongside the notebook. In the app, use Expand code to inspect the query and calculations. Through CLI or MCP, ask your agent to show the cells and execution results.

Check these points:

Compare the same data period and access scope. Different refresh times or database permissions can produce different totals.

If something doesn't match, describe the difference and ask for a correction:

The query includes refunded orders. Exclude them, rerun
the calculation, and show the updated query and total.

Check the revised logic and result. Check and correct an answer covers deeper investigation.

5. Build on what you've checked

Once the result makes sense, take the analysis further:

Break revenue down by week and show the trend.

When you uncover a useful definition or resolve a recurring ambiguity, capture it as context. That vetted knowledge helps the next person asking a related question.

Turn the analysis into a board, share it with your team, or invite teammates to explore it themselves. Check the audience's access before sharing sensitive results.

order_id,paid_at,status,refunded,amount
A,2026-09-01T00:00:00+00:00,paid,false,100
B,2026-09-30T23:59:59+00:00,paid,false,40
C,2026-09-15T12:00:00+00:00,paid,true,30
D,2026-10-01T00:00:00+00:00,paid,false,70
E,2026-08-31T23:59:59+00:00,paid,false,20
F,2026-09-10T12:00:00+00:00,pending,false,10
G,2026-10-01T00:30:00+01:00,paid,false,10