Project: Analyse a Sales Dataset

Why this matters

You now have every piece of a data analysis: load a table, clean it, filter it, group it and summarise it. This project puts them together into a small report, the same shape as a real analysis for a small shop.

The idea

You get a sales log as CSV text (the kind of file a till or spreadsheet exports):

date,product,units,price
2026-01-05,tea,10,2.50
2026-01-20,coffee,4,3.00
2026-02-03,tea,,2.50

One row has no units, so it is not a real sale. The plan:

1. Load and clean: read the CSV, drop rows with no units, convert date to real dates and add a revenue column (units * price).
2. Monthly revenue: total revenue per month, as a dict like {"2026-01": 49.0, ...}.
3. Best products: the top n products by total revenue.
4. Report: three readable lines.

Two tools you have not used yet:

import io
import pandas as pd

df = pd.read_csv(io.StringIO(csv_text))      # read CSV from a string instead of a file
df["date"] = pd.to_datetime(df["date"])       # text -> real dates
df["date"].dt.strftime("%Y-%m")               # "2026-01": the date as text in your format
series.nlargest(3)                            # the 3 biggest values

io.StringIO wraps a string so pandas can read it like a file. That keeps this project self-contained, and the same code works on a real file: pd.read_csv("sales.csv").

Try it

Make a small CSV string, load it with read_csv(io.StringIO(...)), and print df.info() to see the columns and their types.

Your turn

Write these four functions:

• load_sales(csv_text): the cleaned table, with columns date (datetime), product, units, price, revenue, and row numbers from 0.
• revenue_by_month(df): a dict of "YYYY-MM" to total revenue rounded to 2 decimals, in month order.
• best_products(df, n): a list of (product, total revenue rounded to 2) pairs, best first, at most n long.
• report(df): a string of exactly three lines: Total revenue: $<total, 2 decimals>, Best product: <name>, Busiest month: <YYYY-MM> (the month with the most revenue).

Recap

• read_csv + StringIO loads CSV text; to_datetime and .dt handle dates.
• A real analysis is a pipeline: load, clean, derive columns, group, summarise.
• Small functions that each return a simple result are easy to test and reuse.

Going further

Try your functions on a real CSV export (a bank statement, a fitness log). Add a function for the weekday with the most sales (df["date"].dt.day_name()), or chart revenue by month with the next lesson.