InnerVault

Your cart

Your cart is empty. Add a skill and it waits here. See skills you own

Subtotal$0.00

One payment, secure checkout by Whop. Prices in USD.

My skills

Master OpenAI: Turn ChatGPT Into Your Productivity Engine / Research, Files and Data

Data Analysis with ChatGPT

About 15 minutes

ChatGPT can read a spreadsheet, run calculations in Python, and produce tables and charts. You do not need to write code. You do need to check the work. This lesson shows a safe, repeatable method.

How it works

When you upload a spreadsheet and ask a question, ChatGPT can write and run Python code in a sandbox to answer it. OpenAI's help centre notes that this Python environment cannot make external web requests or API calls. So it only knows what is in your file and your prompt.

Step 1: Prepare the data

Clean data gives better answers. Before uploading:

  1. Put one header row at the top, with clear column names: date, product, units, revenue.
  2. Remove merged cells, subtotal rows and notes in the middle of the data.
  3. Use one date format throughout.
  4. Remove personal data you do not need, such as customer emails.
  5. Save as .csv or .xlsx.

Step 2: Ask it to describe the data first

I have uploaded my sales data. Before any analysis:
1. Describe each column and its data type.
2. Tell me the date range and number of rows.
3. List any missing values, duplicates or odd entries.
Do not change the data yet.

This catches problems early. If it misreads a column, correct it now.

Step 3: Ask focused questions

Show total revenue by month as a table, then as a bar chart.
Which 5 products had the biggest drop in units sold between Q2 and Q3?
Show the numbers for both quarters.
What share of revenue comes from the top 10% of customers?

One question per message keeps results easy to check.

Step 4: Check the working

  1. Ask it to show the code or the steps it used.
Explain in plain English the steps you took to calculate that, including any rows you excluded.
  1. Recalculate one figure yourself in your spreadsheet, for example one month's total.
  2. If the numbers differ, ask why. Common causes: duplicates, text values in number columns, or filters you did not intend.

Step 5: Turn findings into decisions

Based on this analysis, give me 3 practical actions for next quarter.
For each, say which number supports it and what could make it wrong.

Step 6: Download outputs

You can ask for a cleaned file or a chart image:

Create a cleaned version of the data with duplicates removed and dates in YYYY-MM-DD format.
Give me the file to download, and list every change you made.

Keep the original file. Never overwrite your source data with an AI-cleaned version without checking it.

A simple monthly routine

  1. Export last month's sales to CSV.
  2. Upload it to a project called "Monthly reporting".
  3. Run a saved prompt:
Using the uploaded file for [month]:
- Total revenue, orders and average order value, compared with the previous month if present.
- Top 5 and bottom 5 products by revenue.
- One chart of daily revenue.
- 3 observations, each tied to a number.
Flag any data problems before the summary.
  1. Check two numbers by hand.
  2. Paste the result into your report.

Common mistakes

  • Uploading messy sheets. Merged headers and subtotals cause wrong totals.
  • Not checking a single figure. One manual check catches most errors.
  • Asking for causes the data cannot show. Sales data shows what happened, not why customers did it.
  • Expecting live data. The sandbox cannot fetch data from the web or your other systems.

Exercise

  1. Export any real table you have: sales, expenses, website traffic.
  2. Clean it using Step 1.
  3. Run the "describe the data" prompt, then two focused questions.
  4. Verify one result by hand and note whether it matched.