Writing better questions

clariBI turns your question into calculations on your data. The clearer the question, the fewer guesses it has to make about columns, periods and totals.

Say the metric, the grouping and the period

Give a time range

Without one, clariBI either uses all the data or picks a window for you.

Good:  Revenue by month from January to June 2026
Good:  Compare Q1 2026 with Q1 2025
Good:  New signups in the last 90 days
Okay:  Recent revenue trend        (a window is picked for you)
Weak:  Revenue                     (no period at all)

Use column names when you know them

Good:  Sum of order_total by product_category
Good:  Average response_time_hours by ticket_type
Okay:  Total sales by product      (it has to guess which column is "sales")

Not sure what the columns are called? Ask "What columns are in the orders file?" or open the data source and look at its preview.

Say count or value

"Orders" can mean the number of orders or their value. "Number of orders" and "order revenue" leave no doubt. The same goes for "average" versus "total", and for rates: a conversion rate or a price should be averaged, not summed.

One question at a time

Instead of:
  Revenue by product, the top customers, their purchase frequency
  and the monthly trend

Ask in turn:
  1. Revenue by product for Q4 2025
  2. Top 10 customers by total spend
  3. How often did those customers buy?
  4. Monthly revenue trend for 2025

Ask for the output you want

Bar chart of revenue by region
Table of customers with more than 5 orders
Show percentages, not absolute numbers
Top 10 by revenue, largest first

Drill down with follow-ups

Follow-ups in the same session keep the context:

1. Total sales by region last quarter
2. Break the top region down by product
3. Which product grew most against the quarter before?

Patterns that work well

• Trend: "[metric] by month for [period]"
• Comparison: "[metric] this quarter against the same quarter last year"
• Ranking: "Top 10 [items] by [metric]" or "Which products have the lowest margin?"
• Share: "What share of revenue comes from each channel?"
• Distribution: "How are order values distributed?"
• Relationship: "Is ad spend correlated with signups?"

"Why" questions get an interpretation of the numbers, not a proven cause. Use them to decide what to look at next.

When the answer looks wrong

1. Check the chart or table: if the numbers are off, the wrong column or period was probably used.
2. Rephrase with explicit column names, the aggregation and the period.
3. Ask for a figure you already know and compare.
4. Limit the session to one data source to find where a number comes from.

Related

• Ask clariBI
• AI limitations
• How charts are chosen