How to use AI to analyze your sales data and tell you which products or services to push this month

You have sales data sitting in a spreadsheet and no idea which products to push next month. This post walks you through using ChatGPT or Julius AI to turn a raw CSV export into a clear, actionable sales picture. Both tools run the analysis for you — no formulas, no pivot tables, no data analyst on retainer.

What you need before you start

Tool name: ChatGPT (GPT-4o or o1) — OpenAI's AI assistant with file upload and data analysis built in. Free tier available; the data analysis features require ChatGPT Plus at $20/month. Check current pricing before you start — plans change.

Tool name: Julius AI — a data analysis tool built specifically for non-technical users. It reads your spreadsheet, runs calculations, and generates charts through a conversational interface. Paid plans start around $20–$25/month as of writing. Check their site for current tiers.

Time required: 45–90 minutes for your first analysis. Under 30 minutes once you have a routine.

Skill level: No coding required. Basic comfort with exporting a file from your POS or ecommerce platform is enough.

---

Preparing Your Data: The Secret to Getting Useful AI Answers

Garbage in, garbage out. This is especially true with AI — it will analyze whatever you give it, including noise.

Before you upload anything, your CSV needs to be clean. Thirty minutes here saves you from acting on wrong conclusions.

1. Open your sales export from Shopify, Square, or QuickBooks. For Shopify users, go to Analytics > Reports > Sales by product and export to CSV.

2. Remove any rows where the order status is "refunded," "returned," or "cancelled." You want completed sales only. You should see a cleaner row count — often 5–15% fewer rows than your original export.

3. Delete columns containing customer names, email addresses, or phone numbers. You do not need this data for trend analysis. You should have columns like: date, product name, SKU, quantity sold, revenue, and category.

4. Add a "Month" column if your date field is day-level. In Excel or Google Sheets, use a simple formula to extract the month from the date column. You should see a clean month label (e.g., "2024-11") next to each row.

5. Save the file as a CSV. Name it something descriptive: sales_clean_nov2024.csv. You should have one file, under 50MB, with no blank rows at the top.

Removing refunds matters because AI tools calculate totals literally. A $400 return sitting next to a $400 sale looks like $800 in revenue. It is not.

---

Step-by-Step: How to Use AI to Analyze Sales Data for Your Small Business

These steps work for either tool. Julius AI is more visual and forgiving for first-timers. ChatGPT gives you more control if you write specific prompts for your chatgpt sales data analysis.

If you're using Julius AI:

1. Go to julius.ai and create an account. You should land on a clean chat interface.

2. Click the upload icon and select your cleaned CSV file. Julius reads the columns automatically. You should see a confirmation message listing your column names.

3. Type your first analysis request in plain language. You do not need to use technical terms. You should see Julius generate a response plus a chart within 30–60 seconds.

4. Review the chart it produces. If it misread a column, tell it in plain language: "The 'revenue' column is in dollars, not units." You should see it recalculate immediately.

If you're using ChatGPT:

1. Open ChatGPT and start a new conversation. Make sure you are using GPT-4o — the model selector is at or near the top of the screen.

2. Click the paperclip icon and upload your CSV. You should see the filename appear in the chat input.

3. Type your prompt before hitting send. Be specific. Vague prompts produce vague answers.

"I've uploaded a CSV of my product sales from the last 6 months. Each row is one order line. Columns are: date, product name, SKU, quantity sold, and revenue. Please analyze this and tell me: (1) which 5 products had the highest total revenue, (2) which 5 products sold the most units, and (3) which products show an upward sales trend over the last 3 months."

4. Read the response carefully. If it gives you numbers that don't match your gut, ask it to show its calculation. You should see the actual math, which lets you verify it.

ChatGPT runs Python in the background to do this analysis. It is not guessing — it is executing code on your actual file. That said, complex multi-step calculations benefit from breaking your question into separate prompts.

---

The 3 Questions You Must Ask Your AI to Get Actionable Insights

Most people ask AI for summaries. Summaries are not decisions. These three questions produce decisions.

Question 1: High-velocity, low-return products

"Which products have the highest sales volume AND the fewest refunds or returns? Show me the top 10 sorted by units sold."

This identifies your reliable sellers — the ones worth running promotions on this month.

Question 2: Low-velocity, high-margin products

"Which products have relatively low sales volume but high revenue per unit? List any product where revenue per unit is more than 2x the average across all products."

This surfaces items worth a targeted push. They sell slowly but pay well when they do.

Question 3: Trend direction by product

"Compare each product's sales volume in the most recent month to their 3-month average. Flag any product where last month's sales were more than 20% higher or lower than the 3-month average."

Upward trend items deserve marketing attention now. Downward trend items need investigation before you reorder.

---

Common Pitfalls: When to Ignore the AI's Advice

Symptom: The AI tells you a product is a top performer, but you know it has a high return rate. Cause: You did not remove refunds from the dataset before uploading. Fix: Clean your CSV, remove refund rows, and re-upload. Re-run the same analysis.

Symptom: The revenue totals the AI reports are significantly higher or lower than your actual figures. Cause: The CSV likely includes tax, shipping, or discount columns that the AI is summing incorrectly. Fix: Tell the AI explicitly: "Use only the 'net revenue' column for all revenue calculations. Ignore the 'gross revenue,' 'tax,' and 'shipping' columns."

Symptom: Julius AI or ChatGPT says a product is trending up, but it only sold twice last month versus once the month before. Cause: Percentage growth on very low numbers is statistically meaningless. Fix: Add a minimum threshold to your prompt: "Only flag trends for products that sold at least 10 units per month."

---

Privacy Checklist: How to Protect Your Customer Data

You should not upload customer names, emails, phone numbers, or addresses to any AI tool — even for analysis. You do not need that data to understand sales trends.

Before uploading, confirm your file passes this checklist:

• No customer names
• No email addresses
• No phone numbers
• No shipping addresses
• No payment details or last-four card digits
• No full order IDs that map back to identifiable individuals

If you need to share data with a team or use it in a business context regularly, both ChatGPT Enterprise and Julius AI's business tiers offer zero-data-retention policies. That means your uploaded files are not used to train their models. Check both platforms for their current enterprise terms before committing sensitive business data.

Shopify has its own built-in AI features under Shopify Magic. These run inside your Shopify account and don't require exporting data to a third party. The tradeoff: Shopify Magic analyzes Shopify data only. Julius AI can compare your Shopify sales against your Google Ads spend or any other CSV you upload alongside it.

---

Monthly Workflow: Creating a Sales Analysis Routine

Do this once a month, on the same day. The first of the month works well — last month's data is complete.

1. Export your sales CSV from your platform. Filter for the previous calendar month. Completed orders only.
2. Clean the file: remove refunds, remove PII columns, add a month column if needed.
3. Upload to Julius AI or ChatGPT and run your three standard questions.
4. Copy the AI's output into a running Google Doc or Notion page labeled by month.
5. Make one decision: pick one product to promote this month based on what you found. Write it down.

The decision step is the only reason you are doing any of this. Analysis without a decision is just reading.

---

What to do next

Take your cleaned CSV from last month and run the three questions above before you do anything else. One run through this process will tell you whether your current product focus matches your actual sales data.

Want to connect your sales analysis to your marketing spend? [See how to automate the connection between your ad data and your sales reports](PENDING: connect Google Ads and sales data using AI automation).

---

FAQ

Can I use AI to analyze sales data if I'm on Shopify's free or basic plan? You can export sales CSVs from any Shopify plan, including Basic. The export is typically found under Analytics > Reports. You then upload that file to Julius AI or ChatGPT separately — neither tool requires a Shopify integration. The analysis happens outside of Shopify entirely.

Is Julius AI better than ChatGPT for this kind of sales analysis? For first-time users, Julius AI handles messy CSVs more gracefully and generates charts automatically without specific prompting. ChatGPT gives you more control and handles follow-up questions well, but it benefits from precise, structured prompts. If you want to compare data across multiple sources — like Shopify plus ad spend — Julius AI handles multi-file uploads more fluidly.

What if my CSV has thousands of rows? Will the AI handle it? Both ChatGPT and Julius AI can handle large files for standard summary analysis. If your file is very large, filter it to the time period you actually need before uploading. A 6-month window is sufficient for monthly trend analysis and keeps the file manageable.