ChatGPT for Data Analysis - The Full Guide

I had a subscriber data export sitting in my Downloads folder for three weeks.

8,000 rows. Open rates, click rates, signup dates, content topics, subscriber sources. Everything I needed to understand what was actually working in the newsletter. I kept telling myself I'd get to it.

I knew what "getting to it" meant. Hours in Excel. Pivot tables. Formulas I'd have to look up. A chart that might not even answer the question I actually had.

I uploaded it to ChatGPT instead. I typed: "What topics get the highest open rates and which subscriber sources convert to paid the most?"

Four minutes later I had two charts and a paragraph that answered both questions.

That's what this guide covers.

When you upload a file to ChatGPT, it doesn't just read the data. ChatGPT writes and executes real Python code against your data in a sandboxed environment  using libraries like pandas for analysis and matplotlib for charts. It runs the code, catches errors, fixes them, and returns the output.

You can click View Analysis at the end of any response to see the exact code it ran.

This is what makes it different from asking ChatGPT to "help me with my spreadsheet" without uploading anything. It's not giving you instructions to follow. It's actually doing the analysis and handing you the result.

You don't need to know Python. You don't need to understand the code. You just need to ask the right question.

This capability has improved significantly in the current generation of models. Multi-file merging, interactive charts, and reliable formula generation were all inconsistent in GPT-4. They work properly now.

Who can use this: ChatGPT's data analysis feature is available on ChatGPT Plus ($20/month), Team, and Enterprise plans. Free users cannot upload files for direct analysis. Verify current pricing at chatgpt.com before subscribing  OpenAI adjusts plans periodically.

CSV files are more reliable than Excel files for analysis. Excel files with heavy formatting, formulas, or embedded images can cause parsing failures. If your data is in Excel, export it to CSV first.

In Excel: File → Save As → CSV (UTF-8). In Google Sheets: File → Download → Comma-separated values.

Keep the file under 50MB. Clean column headers. One record per row. That's all it needs.

Become a paying subscriber to get access to this post and lots of other premium content.

The first thing to do with any new file: ask ChatGPT to tell you what's in it. This takes 30 seconds and saves you from spending 20 minutes manually figuring out what you're working with.

Analyze this dataset. Tell me:

1. How many rows and columns

2. What each column contains

3. The date range if there's a date column

4. Any obvious data quality issues (missing values, inconsistent formats, outliers)

5. The three most interesting patterns you notice immediately

This gives you a map before you start asking specific questions. The fifth point  interesting patterns  consistently surfaces things I wouldn't have thought to look for myself.

I use Typeless to voice-dictate analysis requests when I'm working through a large dataset. Faster than typing and easier to be specific about what I actually want to know.

Once you have the summary, ask your real question. The more specific the prompt, the more useful the answer.

For newsletter data:

Which content topics have the highest average open rate? Show the top 5 as a bar chart, sorted from highest to lowest. Include the number of emails sent for each topic.

For sales data:

Show me monthly revenue for the last 12 months as a line chart. Highlight any month where revenue was more than 20% above or below the previous month.

For subscriber or customer data:

Which acquisition source brings in the most subscribers? Which source has the highest conversion to paid? Show both as separate charts so I can compare.

Specify the metric, the filter, the time period, and the chart type. "Show me top 5 products by revenue in Q4 as a bar chart" gets a precise result. "Analyze my sales data" produces something generic.

If you've ever inherited a spreadsheet from someone else, you know what messy data looks like. Dates in three different formats. Duplicates. Empty rows. Text in columns that should be numbers.

Clean this dataset:

1. Remove any duplicate rows

2. Standardize all dates to YYYY-MM-DD format

3. Remove any rows where the [column name] field is empty

4. Trim any extra spaces from text fields

5. Give me a cleaned version I can download

ChatGPT cleans the file and provides a download link for the result. It cannot overwrite your original file  it processes your uploaded copy and generates a new clean file for you to download safely.

A task that takes 45 minutes in Excel takes two minutes here.

I ran this on a course enrollment export last semester. 1,400 rows with inconsistent date formatting across three academic years. Clean file in 90 seconds.

Any analysis can become a chart. You describe what you want, ChatGPT generates it as a downloadable PNG.

Create a bar chart showing [metric] by [category].

- Sort from highest to lowest

- Use a clean, minimal style

- Add value labels on each bar

- Title: "[your chart title]"

For a line chart:

Show [metric] over time as a line chart.

- X-axis: [time column]

- Y-axis: [metric column]

- Add a horizontal dashed line showing

  the average value

- Title: "[your chart title]"

Bar, line, pie, and scatter charts can be shown as interactive charts  select Switch to interactive chart when the option appears. Other chart types return as static images.

Download the PNG from the chat. Paste it directly into your report, presentation, or newsletter.

If you have two spreadsheets that should be connected  a customer list and a transaction history, a subscriber list and an engagement report  uploading both and asking ChatGPT to merge them eliminates one of the most frustrating Excel tasks.

I'm uploading two files:

File 1: [describe what it contains]

File 2: [describe what it contains]

Merge them using [column name] as the shared identifier. Keep all rows from File 1. Add the [specific columns] from File 2. Give me the merged file to download.

ChatGPT handles the merge and returns a clean combined file. No VLOOKUP syntax. No INDEX/MATCH. No broken references when a name is spelled slightly differently in each file.

File upload limits: ChatGPT Plus supports multiple file uploads per conversation, but exact limits vary by plan, model, and account settings. Check OpenAI's current documentation if you're working with many files in one session.

Sessions expire. The analysis environment expires after 30 minutes of inactivity. When it does, your uploaded files disappear and all variables reset. Download every output immediately after it's generated. Don't assume you can come back to the same chat later.

For long analysis sessions: request intermediate CSV downloads every 15 to 20 minutes.

Verify the important numbers. ChatGPT may choose an analysis method or column that doesn't match your intent on the first try. Click View Analysis after any response to see the Python code it ran. If it summed the wrong column or filtered incorrectly, point it out and it will rewrite and re-run. For any decision that matters, cross-check key numbers against your source data.

It can't fetch live data. The Python environment cannot make external web requests or API calls. It only works with what you upload. If you need live data, export it from whatever tool it lives in and upload the export.

Every strong analysis prompt follows the same structure:

[What to analyze] + [How to filter or group] +

[What time period] + [What output format]

Applied:

Show me [open rate] by [content topic], for [the last 6 months], as a [bar chart sorted highest to lowest].

Vague prompts produce vague results. The more you specify up front, the fewer iterations you need.

I go over this formula on my morning walk with Alfie when I'm planning what to ask before I sit down with the data. By the time I open the laptop, the questions are already written.

If this saves you the afternoon you were about to spend in Excel, send it to one person whose data has been sitting in Downloads for longer than it should.

That's all I'm asking :)

See you next week.