Video 301
Analyzing a sales log with AI, from A to Z
Mark's aunt has a 7,341-row Excel log from 3 months at her bubble tea shop. Mark uses an AI that can run code to answer three questions: which drink earns the most, which one is slipping, and whether to open in the morning. No coding needed, just copy the prompts below.
- Pick an AI that can run code (a code interpreter), such as ChatGPT, Gemini or Claude.
- Delete phone numbers and customer details before uploading. The sample file is fake data, so use it freely.
- Upload the file and paste the prompts below one by one, reading each answer before moving on.
First prompt: make the AI describe the file
This is a bubble tea shop's 3-month sales log. Don't analyze yet. Describe the file: which columns it has, how many rows, what each column holds, and list anything that looks unusual.
Check: make the AI show its working
What total revenue did you get, and which rows did you add up? Show me how you calculated it.
From numbers to decisions
Based on the checked results, suggest at most 3 things the shop should do next month. Give each one a reason in numbers, and say what the data can't tell us for sure.
5-step prompt template: reuse it every month
You are a data analysis assistant. The attached file is [shop name]'s sales log for [month/year]. Work through 5 steps in order, reporting the result of each step before starting the next. 1. Describe the file: its columns, row count, what each column holds, and anything that looks unusual. Don't analyze yet. 2. Clean the data: merge different spellings of the same item, turn every price into a number, use one date format, and drop empty rows and any hand-typed total rows. List every change in a table so I can check it. 3. Answer the measurable questions: [which item has the highest total gross profit, using the cost sheet I attached]; [which item lost the most cups month over month]; [the average cups sold per day between 7 and 10 am, weekdays and weekends separately]. Draw the chart type that suits each question. 4. Check the total revenue two different ways and show me the working. If the two results differ, stop and tell me. 5. Write a one-page summary: at most 3 things to do next month, each with a reason in numbers and a confidence level, and say clearly what the data can't settle.
Video 302
Predicting tomorrow's sales: your first machine learning model in Python
The next level after video 301: use the same sample file to build a model that predicts tomorrow's cups in Google Colab. The notebook below runs from top to bottom, and every number in the video comes from it.
- Click "Open in Google Colab" (needs a free Google account, nothing to install).
- Run each cell in turn with Shift + Enter. The first cell downloads the sample file for you, or upload it with the folder icon on the left.
- Read each block of code before running it. If something is unclear, ask an AI to explain it line by line.
Ask an AI to explain code
Explain each line of the Python code below in plain language for a beginner. Point out any line that could make the results wrong. [paste the code here]