25 Claude Prompts for Data Analysis and Business Intelligence

You are a careful data analyst. I uploaded [file name]. One row = [one order / one customer / one day].
Give me: row and column count, each column's data type, missing values per column, unique values per column,
and 5 sample rows. Do not change the data. Show the code you used.

Create a data dictionary for this file as a table: column name, data type, example value,
likely meaning, and any value that looks suspicious. Mark guesses as "assumed".

For each numeric column, show min, max, mean, median and the number of values outside 1.5 x IQR.
List the 10 most extreme rows for [amount] so I can check them.

For each text column with fewer than 50 unique values, list the values and counts.
Flag spelling variants that probably mean the same thing (for example "Mumbai", "mumbai", "Bombay").

Propose cleaning rules for this dataset as a numbered list: the problem, how many rows it affects,
the fix you suggest, and the risk of that fix. Wait for my approval before applying anything.

Write pandas code to detect missing values and fill them:
[age] with the median per [group], [city] with "Unknown", and drop rows where [order_id] is missing.
Print row counts before and after each step.

Find duplicate rows on [order_id]. Show how many there are, 5 examples,
and pandas code that keeps the latest record by [updated_at].

[order_time] is stored in UTC. Write pandas code to convert it to India time (Asia/Kolkata),
add order_date, month and weekday columns, and show 5 rows before and after.

Merge [orders.csv] and [customers.csv] on [customer_id] with a left join.
Report how many orders have no matching customer, and check that the row count did not increase after the merge.

Question: [Which region grew fastest from Q1 to Q3 2026?]
Answer with a small table, the growth % calculation, and the pandas code you used.

Compare [new vs returning customers] on average order value, orders per customer and revenue share.
Present a table and 3 plain-English findings.

Show monthly [revenue] for the last 12 months, the month-over-month change,
and any month that is more than 20% above or below the average. Suggest possible reasons I should check.

List the top 10 and bottom 10 [products] by [revenue] and by [margin].
Flag any product that is in the top 10 for revenue but the bottom 10 for margin.

[Revenue] dropped in [August]. Break the drop down by region, product category and channel,
and show which breakdown explains most of the change.

I need to show [revenue by month and region] to a manager. Suggest the 2 best chart types,
explain why in one line each, and give pandas plot code for the first one.

Design a one-page sales dashboard: 4 KPI cards, 3 charts and 1 table.
For each item give the metric, its exact formula and the filter it should respond to.

Write clear definitions for these KPIs so two people cannot calculate them differently:
[revenue, active customers, repeat rate, average order value].

In Excel, column A has order dates and column D has amounts. Write a formula for total amount
in [March 2026] and explain each part. Then give the same result using a PivotTable.

Suggest 5 new columns I can add in Excel to make this customer file more useful for analysis
(for example tenure in months, order frequency band). Give the formula for each.

I have 12 monthly CSV files with the same columns. Give me Power Query steps to combine them,
remove blank rows, fix data types and add a Month column. Use the Power Query editor menu names.

I have source.csv and target.csv for the same orders. Write pandas code that compares row counts,
total [amount] and counts by [status], then lists order_ids that exist in only one file.

My dashboard shows [revenue = 48,20,000 for March]. Recalculate it from the raw data,
show the filter you used, and list possible reasons if the numbers do not match.

Write 10 data quality checks for this table as SQL queries (not null, unique, valid range,
valid category, referential integrity). Each query should return the failing rows.

Write a 5-line summary of these results for a non-technical manager:
1 headline, 3 findings with numbers, 1 recommended action. No claim that the data does not support.

List every assumption and data limitation in this analysis that the reader should know,
in order of importance.