"DataFrames: load and explore"

01The problem

The sales director has sent Bisi three files from Kolanut's system: orders.csv, customers.csv and products.csv, and asked "what's in here?" before anyone builds a report on them.

That question comes first on every analysis. How many rows? Which columns, and what type is each? Are there gaps? What are typical values, and are there any odd ones? Ten minutes of exploring now saves you from a wrong answer later. In pandas, each of those questions is one short line of code.

02The concept

pandas and the DataFrame

pandas is the Python library for tables. Its main object is the DataFrame: rows and named columns, like a sheet in Excel. Each column is a Series, one column of values that all share a type. Everyone imports pandas with the short name pd:

import pandas as pd

Loading a CSV

pd.read_csv() reads a file from a web address or from your computer:

base = "https://academy.cloudtechanalytics.com/datasets/sales/"
orders = pd.read_csv(base + "orders.csv")
customers = pd.read_csv(base + "customers.csv")
products = pd.read_csv(base + "products.csv")

The first questions, in code

Selecting columns

• One column, as a Series: orders["quantity"].
• Several columns, as a DataFrame: orders[["order_date", "quantity"]]. Note the double brackets: the inner pair is a list of names.

Series have their own methods: .sum(), .mean(), .min(), .max(), .median(), .count().

Types in pandas

info() shows types with pandas names: int64 (whole numbers), float64 (decimals), object (usually text), datetime64 (dates), bool. A date column that shows as object is being treated as text: it'll sort, but you can't take the month out of it or do date maths. You'll fix that in lesson 5.

03Example

Load the three files and look at the orders:

import pandas as pd

base = "https://academy.cloudtechanalytics.com/datasets/sales/"
orders = pd.read_csv(base + "orders.csv")
customers = pd.read_csv(base + "customers.csv")
products = pd.read_csv(base + "products.csv")

print(orders.shape)
orders.head()

(4266, 7)
   order_id  order_date  customer_id  product_id  quantity  unit_price  discount_pct
0     10001  2025-01-01           27           3        14       18600             0
1     10002  2025-01-01           56           1         7       13200             0
2     10003  2025-01-01           37           2         4        3600             0
3     10004  2025-01-01           26           7        28        6000             5
4     10005  2025-01-01           27           8        19       10500             5

Each row is one order line: one product on one order. The numbers on the left (0, 1, 2…) are the index, pandas' row labels.

orders.info()

info() reports 4,266 non-null values in every column, so nothing is missing, and shows order_date as object: text, for now.

orders.describe().round(1)

From describe(): quantities run from 1 to 30 cartons with an average of 13.8, unit prices from ₦3,600 to ₦24,600, and discounts are 0, 5 or 10%.

04Walkthrough

1. Open a new Colab notebook and run the Example's first cell to load the three files.
2. Run customers.head() and products (a table this small you can just display). Note how they link to orders: customer_id and product_id.
3. Check the sizes: customers.shape is (90, 8) and products.shape is (16, 4).
4. Run orders["discount_pct"].value_counts(): 2,519 lines had no discount, 1,222 had 5% and 525 had 10%.
5. Add normalize=True to get shares instead: orders["discount_pct"].value_counts(normalize=True). About 59% of lines had no discount.
6. How many different customers ordered? orders["customer_id"].nunique() gives 90: every customer has at least one order.
7. Select two columns with double brackets, orders[["order_date", "quantity"]].head(), and compare it with the single-bracket orders["quantity"].head(). One is a table, the other a single column.
8. Add a text cell summarising the dataset in three sentences: what one row is, the date range, and anything you'd want to check.

05Practice

06Challenge

07Check your understanding