How to Ask Good Questions of Your Data, With AI Help

"Can you look at the sales data?" is one of the most common requests in any office, and one of the least useful. Without a clear question, analysis turns into hours of making tables that nobody uses. Good analysis starts before you open the spreadsheet: with the right question.

Start from the decision

Every useful question is linked to a decision someone will make. Ask:

Who will use the answer?
What decision will they make with it?
By when?

Vague request	Decision behind it	Good question
"Look at sales"	Which products to stock more of before Diwali	Which 10 products grew most in Sep–Oct compared with the same months last year?
"Check expenses"	Where to cut costs next quarter	Which 5 expense categories grew faster than sales in the last 6 months?
"How are customers doing?"	Who should the sales team call this week	Which customers bought every month last year but have not ordered in 60 days?

Let AI brainstorm, then you choose

I am a [role] at a [type of business]. My manager asked me to "look at
our [sales/expenses/orders] data". The decision coming up is [decision,
if known]. My data has these columns: [list].
Suggest 10 specific questions this data could answer that would help
with that decision. For each, say which columns it needs.

Pick two or three questions that matter most. More questions do not mean better analysis.

Make each question answerable

A good analysis question is:

Specific: which measure (sales amount, number of orders, margin), which group (product, customer, region) and which period
Comparable: compared with what? Last month, the same month last year, a target, another region
Answerable with your data: if you do not have the column, you cannot answer it

Ask the AI to sharpen your question:

Make this question more specific and measurable, and tell me what it
should be compared with: "Are our sales getting better?"

A sharpened version might be: "How did total monthly sales from April to October compare with the same months last year, by product category?"

Check what the data can really tell you

Before analysing, ask:

Given these columns [list] and this question [question], what are the
limits of what my data can tell me? What could make the answer
misleading?

Typical limits: returns not recorded, prices changed during the period, a new branch opened mid-year, or data missing for some months. Knowing these up front saves embarrassment later.

Write a one-line analysis brief

Before starting, write one line:

Question: Which 10 products grew most in Sep–Oct vs last year? For: the owner, to decide Diwali stock. By: Friday. Data: sales export Apr 2025 – Oct 2026.

This keeps you focused and makes it easy to report back.

Practice task

Take one real request you received recently (or imagine one). Write down the decision behind it, use AI to brainstorm ten questions, choose the best two, sharpen them, and write your one-line brief.

Key takeaways

Every good data question is tied to a decision, a person and a deadline.
Use AI to brainstorm questions, then choose the two or three that matter most.
Make questions specific, comparable and answerable with your columns.
Write a one-line brief before you open the spreadsheet.