This lesson is being recorded
Every video of this course is being uploaded to my YouTube channel. Subscribe so you do not miss it.
Bruno Ashford @brunoashfordSubscribeModule 3 · Lesson 3.3 · 13 min
Making Sense of Messy Spreadsheets
Turn a messy real-world spreadsheet into clean data and plain-English answers, without touching the original file.
Use ← → to move. The marker highlights text and the pencil draws on the slide; navigation locks while you annotate. Undo removes the last mark, Esc exits.
1Put the file in the folder and protect the original
Download the practice file 03-sample-messy-sales.csv into your Company/sales folder. Tell Claude the original is read-only. Every cleaning job should produce a new file so you can always go back.
The file sales/03-sample-messy-sales.csv is the original. Never edit it. Any cleaned version goes into a new file with -clean in the name.2Ask for a mess report first
Before cleaning anything, get a list of every problem Claude can find. This is your chance to make business decisions, like whether a refund counts as a sale.
Read sales/03-sample-messy-sales.csv and give me a mess report: inconsistent dates, duplicate rows, customer names spelled differently, missing values, amounts stored as text, and anything else odd. Do not fix anything yet. For each issue, suggest how you would handle it.3Decide the rules, then let it clean
Answer its questions in business terms. You decide what a duplicate is and how refunds are treated. Claude does the tedious part and writes the clean copy.
Go ahead with these rules: dates as YYYY-MM-DD, merge customer names that are clearly the same company, remove exact duplicates, keep refunds as negative amounts, and flag any row you are unsure about in a new column called check. Save it as sales/03-sample-messy-sales-clean.csv.4Ask business questions, not spreadsheet questions
You do not need to know formulas. Ask the question you would ask your finance person, and ask for the answer in a small table with the method explained.
Using the clean file: what was revenue by month, who are the top 5 customers by revenue, which sales rep has the highest average deal size, and what share of revenue came from refunds? Show small tables and explain in one line how you calculated each.5Reconcile before you trust
The clean total should match the original total minus what you chose to remove. Ask for that reconciliation every time. If it does not add up, you have found either a bug or a business insight.
Reconcile: what is the total in the original file, what is the total in the clean file, and exactly which rows explain the difference?Every business I have ever looked inside has the same spreadsheet: the one three people edit, with dates in four formats, a customer spelled five ways and a column called "notes 2". You know the answers are in there. You just never have the two hours to dig them out.
This is where Claude Code earns its keep fast. It can read the file, find the mess, clean a copy and answer your questions in minutes. In this lesson you will do it on the practice file included with this module, then on your own.
Key points
- Never let Claude edit the original file. Always a cleaned copy.
- Get a mess report before any cleaning.
- You set the business rules. Claude does the tedious work.
- Ask business questions in plain English.
- Always reconcile totals between original and clean data.
Avoid
- Letting the agent overwrite the only copy of your data.
- Accepting its cleaning choices without deciding the business rules yourself.
- Asking for formulas when you really want an answer.
- Skipping the reconciliation because the numbers look reasonable.
Priya sells homeware online and exports orders from two platforms into one spreadsheet. The mess report found the same customer under three email variants, 42 test orders from her own checkout, and currency symbols mixed into the amount column. After cleaning, the top 20 customers turned out to drive 38 percent of revenue, a figure she had guessed at 15 percent. She launched a VIP email the same week.