Excel is not a lesser tool
There's a snobbery in analytics that says real work happens in Python. I've watched a well-built spreadsheet change how a business runs.
There is a quiet snobbery in analytics: real work happens in Python, and spreadsheets are what you graduate from.
I don't buy it. Badly built spreadsheets are a lesser tool. A well-built one is infrastructure.
The problem, in a real shop
Small business owners keep books by hand. Every row of a sales log is an opportunity for a slipped digit. Worse, the arithmetic is retrospective — the owner finds out what their margin was long after the money moved and the decisions were made.
Ask a shop owner which item earns them the most and you'll often get a guess. Not because they're careless. Because the answer requires an evening with a calculator that they don't have.
What I built
A daily sales log where the owner enters three things — quantity, cost price, selling price — and the workbook does the rest:
- Total cost, total revenue, and net profit computed per item, the moment the row is entered
- A summary matrix at the top that aggregates everything into live totals
- A layout designed for real daily inputs, from staple goods to bulk commodities
Across the month I built it on, that came to ₦153,500 in revenue and ₦50,750 in net profit — a 33% margin the owner can now see without doing a single calculation.
Why this counts as analytics
Three reasons.
It removes an error class entirely. Once data is entered, no human does arithmetic. The mistakes that plague manual bookkeeping simply cannot occur.
It converts hindsight into foresight. The most profitable items are visible now, not at month-end. That changes what the owner stocks tomorrow.
It gets used. This is the part the Python snobbery misses. A model nobody runs is worth nothing. A shop owner will open Excel. They will not open a Jupyter notebook. The best tool is the one that survives contact with the person who has to use it every day.
The principle
Choose the tool that fits the user, not the one that impresses your peers. Sometimes that's a regression in R. Sometimes it's a ruthlessly well-built spreadsheet.
Knowing which is half the job.
Next post
The model that was too perfect