At most one email a week. Unfollow anytime with one click.
Transcript
The full narration of this video. The current sentence is highlighted while the video plays; click any sentence to jump there.
Show transcript
Introduction
Hi, and welcome back. In the previous episode we read PDF documents. Now we move on to numbers. We'll learn to read Excel and CSV data with Claude: preparing tidy data, uploading and asking questions, cleaning messy data, building a visual summary, and checking the numbers ourselves.
Step 1: Prepare tidy data
Laras records every Sunset Coffee order in a spreadsheet. The data is neatly stored, but rarely read, because she never has time to build formulas and charts. Claude can help turn those rows of numbers into answers: which product sells best, what the total is, and which customers buy most often. There are two common file types. Excel files, and CSV files, which are simple table files you can save from Excel or Google Sheets. Both can be attached to Claude. Before uploading, tidy up the data. First, the column headings are in the first row, for example date, customer, product, quantity, and total. Second, one row per transaction. Don't combine several orders in one cell, because it becomes hard to count. Third, numbers are written as numbers. Write eighteen thousand as digits, not as a shorthand like eighteen k. Fourth, delete personal data columns such as customers' phone numbers, addresses, and emails. Claude doesn't need them to calculate sales. The safest way: make a copy of the file just for analysis, then delete the personal columns in that copy. Your original file stays intact on your computer.
Step 2: Upload and ask
Here's a slice of Sunset Coffee's order data for the first three days of October. The phone number column has already been deleted. In the first row, Rina buys three iced coffee cups. Budi buys two cold brews, Sari one liter of iced coffee, and Andi one gift hamper. Look at Putri's row. The product is written as ic cup, a shorthand Dimas uses when the shop is busy. Then there are two rows for Joko that are exactly the same. Did Joko really order twice, or did Dimas record it twice by accident? Laras attaches the CSV file and asks Claude to check for odd data first, and only then calculate. Claude reads all seven rows of data. Claude finds two things: Joko's duplicate row, and ic cup, which is probably the same as the iced coffee cup. Notice that Claude asks first instead of deciding on its own. Laras is the one who knows her shop. She confirms that Joko's second row is indeed a duplicate, and that ic cup is the iced coffee cup. After cleaning, Claude counts six transactions. Cold brew brings in the most money, one hundred sixty-five thousand rupiah, while the iced coffee cup sells the most units, seven cups. The total is five hundred twenty-six thousand rupiah. Imagine if that duplicate row hadn't been deleted. Total sales would be fifty-five thousand higher than reality, and Laras could misjudge her business. The analysis depends heavily on our question. Analyze this data is too vague. Better: calculate the total per product and sort from largest. How are my sales is vague too. It's clearer to ask for a comparison, for example the first and second weeks of October. Who's my best customer can be read in many ways. Set the measure, for example: who bought more than two times this month? The rule is similar to writing prompts: the more specific the measure, the more useful the answer. If you're not sure what to ask, have Claude suggest five questions worth asking about your data.
Step 3: Common data problems
The problems we just found are very common. The first is inconsistent product names. One product can be written three different ways by different people. Second, duplicate rows, from an order recorded twice, for example when a customer resends a message. Third, mixed date formats. Some people write zero one slash ten, others write October first. Fourth, numbers written as text, like eighteen k. The spreadsheet can't add them up. Fifth, empty cells, such as a quantity or total someone forgot to fill in. Claude can detect all five problems. But the decision on how to fix them stays in your hands, because only you know what actually happened. Even better, fix how the data is recorded. Laras now uses a product dropdown list in the spreadsheet, so Dimas can no longer type ic cup. Let's sum up the workflow. First, tidy the data and delete personal data columns. Second, upload the Excel or CSV file to Claude. Third, ask Claude to check for odd data before calculating, then decide the fixes yourself. Fourth, ask specific questions, and ask for the results as a table or a chart. Fifth, match the important numbers against spreadsheet formulas. We'll look at the visual summary and this matching step now.
Step 4: A visual summary as an Artifact
The data is clean. Now Laras asks for a visual summary: total, number of transactions, average per transaction, a chart by product, and one finding. The right panel shows number cards at the top, a bar chart in the middle, and one finding at the bottom. The average spend per transaction is about eighty-seven thousand rupiah. The finding is interesting: the iced coffee cup is bought most often, but cold brew brings in the most money. Laras then asks for one suggestion to try this week. Claude adds a suggestion, and honestly reminds her that it's based on only three days of data. For a more confident decision, Laras needs a full month of data. Laras can share a summary like this with Dimas for their weekly meeting. No need to open the spreadsheet; just look at the number cards and the chart.
Step 5: Check the numbers yourself
The last step is checking the numbers yourself. Laras copies the per-product table into the spreadsheet. Then she adds up the total column with a SUM formula. The result is five hundred twenty-six thousand, the same as Claude's calculation. A SUM formula never makes an addition mistake. So for any number you'll report or use to make a decision, always match it against a spreadsheet formula. This habit sounds trivial, but it's what separates a trusted report from a doubted one. Claude speeds up the calculation, and the spreadsheet formula makes sure the result is right. Keep these five habits. First, match the total against a SUM formula, just as Laras did. Second, check the row count before and after cleaning. Earlier, seven rows became six, because one duplicate row was deleted. Any difference must be explainable. Third, ask Claude to explain how it calculated, step by step. If a step looks odd, you can correct it right away. Fourth, check two or three random rows against the original data. Fifth, make sure the uploaded file has no customer personal data. File reading and analysis capabilities can differ by plan and settings, so check Claude's official documentation. The screens in this video are simulations.
Done!
We've learned to prepare tidy data without personal information, upload a CSV and have Claude check for odd data, clean the data, build a visual summary, and match the results against a SUM formula. In the next episode we read a different kind of file: images, photos, and screenshots, from a photo of a shopping receipt to a screenshot of an error message. See you there!
Practice exercises
Try these on your own after watching to lock in what you learned.
Tidy up one of your data files: column headings in the first row, one row per transaction
Upload that file and ask Claude to check for odd data before calculating
Ask for a visual summary, then match the total against a SUM formula in your spreadsheet
Ask the class material
Type a question or keyword; we find the part of this episode that explains it and jump the video there.
My notes
Notes are saved with the current video time and are only visible to you.
Questions & answers
Stuck on a step? Ask here; the instructor and other learners can answer.
No questions yet.
Your instructor
Cerdas Inovasi Digital Team
Software developers, BIM & 3D engineers, and AI practitioners
52 video classes1,276 episodes
We build software, BIM models, games and AI automations for clients every day, and turn that hands-on work into step-by-step classes: every step is shown on screen and comes with a PDF guide you can keep.
