Writing better questions for AI analytics: 40 examples
A vague question gets a vague answer. These 40 examples show the same question asked badly and asked well, across sales, marketing, customer success and finance, and the five patterns behind every good version.
When you ask an AI analytics tool a question, it has to work out four things: which metric you mean, how to group it, which filters apply (above all the period), and what to compare it with. A vague question leaves some of these open, and the tool fills the gaps with guesses. A specific question closes them.
Below are 40 questions, each asked the weak way and the better way, grouped by team. Adapt the names to your own columns and tools. Each better version assumes the data exists in a connected source; if it does not, a good tool says which fields are missing rather than guessing.
Sales and revenue
| # | Weak | Better |
|---|---|---|
| 1 | How are sales? | What was total revenue from paid orders in Q2 2026, compared with Q2 2025? |
| 2 | Show me the pipeline. | Show open deals by stage, with count and total value per stage, for deals expected to close this quarter. |
| 3 | Who is our best salesperson? | Rank reps by closed-won revenue in the last 90 days, with deal count and average deal size. |
| 4 | What is our win rate? | Of deals closed in the last 6 months, what share were won, by month? |
| 5 | Which products are doing well? | Show revenue by product category for the last 12 months with month-over-month change. |
| 6 | How fast do we close? | Average days from deal created to closed-won, for deals closed in the last 6 months, by deal size band. |
| 7 | Are we hitting our numbers? | Each rep's closed-won revenue this quarter against their quota, sorted by gap to quota. |
| 8 | Where is revenue coming from? | Split this quarter's revenue into new customers and existing customers, compared with last quarter. |
| 9 | Why are we losing deals? | For deals lost in the last 90 days, count by loss reason, with average deal value per reason. |
| 10 | Which customers matter most? | Top 10 customers by revenue this year, and their share of total revenue. |
Example 4 matters more than it looks: win rate depends on the denominator. Counting open deals, or all leads, gives a different number. Example 9 turns a "why" question, which data cannot answer, into a "what" question a person can reason from.
Marketing
| # | Weak | Better |
|---|---|---|
| 11 | How is marketing doing? | Leads by channel for the last 3 months, with cost per lead for each channel. |
| 12 | Was the campaign worth it? | For the spring campaign, total spend, conversions and cost per conversion, by ad set. |
| 13 | How is the website doing? | Weekly sessions and sign-up conversion rate for the last 8 weeks. |
| 14 | Which content is popular? | Top 10 pages by views in the last 90 days, with engagement rate for each. |
| 15 | How are our emails doing? | Open rate, click rate and unsubscribe rate for each campaign sent in the last 60 days. |
| 16 | Are the leads any good? | Of leads created last quarter, what share became customers within 60 days, by lead source? |
| 17 | Are we spending too much on ads? | Monthly spend on Google Ads and Meta Ads next to conversions from each, with cost per conversion. |
| 18 | How is our SEO? | Clicks, impressions and average position from Google search by month, last 12 months. |
| 19 | What does a customer cost us? | Total sales and marketing spend divided by new customers, by quarter, last 4 quarters. |
| 20 | Show me the funnel. | Conversion rate between each stage (visit, sign-up, trial, paid) for the last 6 months. |
For 17, both ad sources (Google Ads and Meta Ads) need to be in the session. For 18, note that average position is a rate: it should be averaged, not summed. Example 19 is customer acquisition cost; the CAC post covers the variants.
Customer success and support
| # | Weak | Better |
|---|---|---|
| 21 | How is churn? | Monthly customer churn rate for the last 12 months, and churned customers by plan. |
| 22 | Which customers are at risk? | Customers whose logins in the last 30 days fell below a quarter of their 6-month monthly average. |
| 23 | What do customers think of us? | Average satisfaction score by month for the last 12 months, by customer tenure band. |
| 24 | How is support doing? | Weekly ticket volume, average first response time and average resolution time, last 3 months, by priority. |
| 25 | Are people using the new feature? | Share of active users who used the feature at least once, by week since launch. |
| 26 | How is onboarding going? | Of customers who signed up in the last 90 days, what share finished each onboarding step within 14 days? |
| 27 | Are customers upgrading? | Number of upgrades last quarter and the added monthly revenue, by plan upgraded from. |
| 28 | Are we keeping customers? | For each monthly sign-up cohort this year, the share still active after 30, 60 and 90 days. |
| 29 | What are customers complaining about? | Tickets by category for the last 60 days, and whether each top category is rising week over week. |
| 30 | Are big accounts engaged? | For the largest 20 accounts, active users as a share of seats, and the date of the last admin login. |
Question 22 only works if login events are in your data; question 28 is a cohort question, which cohort analysis explains.
Finance and operations
| # | Weak | Better |
|---|---|---|
| 31 | How is cash flow? | Monthly inflows, outflows and ending balance for the last 6 months. |
| 32 | Are we overspending? | Actual spend against budget by department this quarter, with the variance in percent. |
| 33 | What is our MRR? | MRR at the end of each month this year, and the change from month to month. |
| 34 | What are our margins? | Gross margin by product line for each of the last 4 quarters. |
| 35 | Are we efficient? | Revenue per employee by quarter for the last 4 quarters. |
| 36 | How are collections? | Unpaid invoices by age: 0 to 30, 31 to 60, 61 to 90 and over 90 days overdue. |
| 37 | Where is the money going? | Top 20 suppliers by spend in the last 12 months, with the change from the year before. |
| 38 | How long will our money last? | Average monthly net burn over the last 6 months, and current cash divided by it. |
| 39 | Do we make money on each customer? | Average revenue per customer and gross margin per customer, by quarter. |
| 40 | Are we meeting our SLAs? | Share of tickets resolved within target, by priority and week, for the last 8 weeks. |
Question 38 is runway: with 300,000 in cash and an average net burn of 25,000 a month, runway is 300,000 / 25,000 = 12 months. For question 34, say how gross margin is calculated if your data has more than one cost column: (revenue minus cost of goods sold) divided by revenue.
The five patterns behind every better question
1. Name the metric precisely. "Revenue" beats "sales". "Revenue from paid orders" beats "revenue". Use your own column names when there is any doubt: "sum of amount where status is paid".
2. Give the period. "Q2 2026", "the last 90 days", "each month this year". "Recently" is not a period.
3. Add a comparison. Last period, same period last year, a quota. A number alone cannot tell you whether it is good.
4. Say how to break it down. By product, channel, rep, plan or week. Without a grouping you get one number.
5. Ask for evidence, not explanations. "Why is churn up?" invites a story. "Churn by plan, tenure and cancellation reason for the last 6 months" gives you the facts to find the reason yourself.
Two more habits help. Ask one thing at a time and use follow-ups ("break that down by region", "only the top five"); the session remembers the previous answer. And ask only about data you have connected: an analytics tool that works on your data cannot look up industry benchmarks or a competitor's revenue.
Using these in clariBI
Open Ask clariBI, start a session with the sources the question needs, and paste one of the better versions with your own names. clariBI plans the analysis, computes the numbers with regular code on your data, and writes the answer with a chart; any narrated number the data cannot back is removed. Click a chart point to see the dataset, date range and rows behind a value. Quick Suggestions offers questions based on your data if you are unsure where to start. See your first question for the basics, the help center's 30 example questions that get good results for more examples with the data each needs, and the writing better questions docs for phrasing tips. How natural language analytics works explains why specific questions work better.
Pick the weak question you asked most recently, rewrite it with the five patterns, and ask both versions. The difference in the answers is the best argument for the habit.