ecommerce

Suppose you are working with an E-Commerce platform and you need to analyze and come up with many metrics and insights.

import numpy as np
import pandas as pd

df = pd.read_csv('new_retail_data.csv')
df

Out[4]:

	Customer_ID	Name	City	State	Country	Age	Gender	Income	Customer_Segment	Year	Month	Total_Purchases	Amount	Product_Category	Product_Brand	Product_Type	Feedback	Payment_Method	Ratings	products
0	37249.0	Michelle Harrington	Dortmund	Berlin	Germany	21.0	Male	Low	Regular	2023.0	September	3.0	108.028757	Clothing	Nike	Shorts	Excellent	Debit Card	5.0	Cycling shorts
1	69749.0	Kelsey Hill	Nottingham	England	UK	19.0	Female	Low	Premium	2023.0	December	2.0	403.353907	Electronics	Samsung	Tablet	Excellent	Credit Card	4.0	Lenovo Tab
2	30192.0	Scott Jensen	Geelong	New South Wales	Australia	48.0	Male	Low	Regular	2023.0	April	3.0	354.477600	Books	Penguin Books	Children's	Average	Credit Card	2.0	Sports equipment
3	62101.0	Joseph Miller	Edmonton	Ontario	Canada	56.0	Male	High	Premium	2023.0	May	7.0	352.407717	Home Decor	Home Depot	Tools	Excellent	PayPal	4.0	Utility knife
4	27901.0	Debra Coleman	Bristol	England	UK	22.0	Male	Low	Premium	2024.0	January	2.0	124.276524	Grocery	Nestle	Chocolate	Bad	Cash	1.0	Chocolate cookies
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
302005	12104.0	Meagan Ellis	Townsville	New South Wales	Australia	31.0	Male	Medium	Regular	2024.0	January	5.0	194.792597	Books	Penguin Books	Fiction	Bad	Cash	1.0	Historical fiction
302006	69772.0	Mathew Beck	Hanover	Berlin	Germany	35.0	Female	Low	New	2023.0	December	1.0	285.137301	Electronics	Apple	Laptop	Excellent	Cash	5.0	LG Gram
302007	28449.0	Daniel Lee	Brighton	England	UK	41.0	Male	Low	Premium	2024.0	February	3.0	60.701761	Clothing	Adidas	Jacket	Average	Cash	2.0	Parka
302008	45477.0	Patrick Wilson	Halifax	Ontario	Canada	41.0	Male	Medium	New	2023.0	September	1.0	120.834784	Home Decor	IKEA	Furniture	Good	Cash	4.0	TV stand
302009	53626.0	Dustin Merritt	Tucson	West Virginia	USA	28.0	Female	Low	Premium	2024.0	January	7.0	340.319059	Home Decor	Home Depot	Decorations	Average	Cash	2.0	Clocks

302010 rows × 20 columns

To work on the E-Commerce dataset, we need to understand two things that we will use in our analysis.

1

To find Unique Values in a single Column

As you can see there are 302,010 rows in the dataset, but customers repeat. To find unique customers use unique or nunique.

df['Customer_ID'].unique()

Out[5]:

array([37249., 69749., 30192., ..., 49840., 58286., 29845.], shape=(86767,))

df['Customer_ID'].nunique()

Out[6]:

86766

There are 86,766 unique customers in this dataset.

A real example: Suppose you want to find Country-wise Number of Customers. If you count rows per country you will double-count customers who appear multiple times. Use nunique to count unique customers per country.

Wrong approach (counts rows, not unique customers):

country_customer_count = df.groupby('Country').agg(
    customer_count = ('Customer_ID', 'count')
)
country_customer_count

Out[7]:

Country	customer_count
Australia	45282
Canada	45252
Germany	52774
UK	63002
USA	95122

Right approach (unique customers per country):

country_customer_count = df.groupby('Country').agg(
    customer_count = ('Customer_ID', 'nunique')
)
country_customer_count

Out[8]:

Country	customer_count
Australia	35430
Canada	35169
Germany	39625
UK	45034
USA	58466

2

Create an Age Group Column

To bucket ages into groups (10–20, 20–30, etc.) use pd.cut:

df['Age_Group'] = pd.cut(df['Age'], bins=np.arange(10, 101, 10))
df

Sample output (Age_Group column shown):

	Customer_ID	Age	...	products	Age_Group
0	37249.0	21.0	...	Cycling shorts	(20, 30]
1	69749.0	19.0	...	Lenovo Tab	(10, 20]
2	30192.0	48.0	...	Sports equipment	(40, 50]
3	62101.0	56.0	...	Utility knife	(50, 60]
4	27901.0	22.0	...	Chocolate cookies	(20, 30]
...	...	...	...	...	...
302005	12104.0	31.0	...	Historical fiction	(30, 40]
302006	69772.0	35.0	...	LG Gram	(30, 40]
302007	28449.0	41.0	...	Parka	(40, 50]
302008	45477.0	41.0	...	TV stand	(40, 50]
302009	53626.0	28.0	...	Clocks	(20, 30]

---

Which customer segment generates the highest total revenue?

Create a Total_Amount column and aggregate by Customer_Segment.

df['Total_Amount'] = df['Amount'] * df['Total_Purchases']

customer_segment_revenue = df.groupby('Customer_Segment').agg(
    revenue = ('Total_Amount', 'sum')
)
customer_segment_revenue = customer_segment_revenue.sort_values('revenue', ascending=False)
customer_segment_revenue

Out[10]:

Customer_Segment	revenue
Regular	1.997659e+08
New	1.244222e+08
Premium	8.757099e+07

Convert to percentage:

total_revenue = customer_segment_revenue['revenue'].sum()
customer_segment_revenue['per'] = customer_segment_revenue['revenue'] / total_revenue * 100
customer_segment_revenue

Out[11]:

Customer_Segment	revenue	per
Regular	1.997659e+08	48.515233
New	1.244222e+08	30.217239
Premium	8.757099e+07	21.267528

Insight: 49% of revenue comes from Regular customers.

---

Which customer segment generates the highest revenue per user (ARPC)?

ARPC = Total Revenue / Unique Customers.

Overall ARPC:

total_revenue = df['Total_Amount'].sum()
unique_Customer = df['Customer_ID'].nunique()
arpc = total_revenue / unique_Customer
arpc

Out[12]:

np.float64(4749.413929779314)

ARPC per customer segment:

cdf = df.groupby('Customer_Segment').agg(
    total_revenue = ('Total_Amount', 'sum'),
    unique_Customer = ('Customer_ID', 'nunique')
)
cdf['arpc'] = cdf['total_revenue'] / cdf['unique_Customer']
cdf = cdf.sort_values('arpc', ascending=False)
cdf

Out[14]:

Customer_Segment	total_revenue	unique_Customer	arpc
Regular	1.997659e+08	72209	2766.496020
New	1.244222e+08	57109	2178.680166
Premium	8.757099e+07	45549	1922.566707

Insight: Highest ARPC is Regular (≈2,766) and lowest is Premium (≈1,923).

---

Which product category has the highest sales volume?

Sales volume = total units purchased (sum of Total_Purchases).

product_sales_vol = df.groupby('Product_Category').agg(
    sale_volume = ('Total_Purchases', 'sum')
)
product_sales_vol = product_sales_vol.sort_values('sale_volume', ascending=False)
product_sales_vol

Out[15]:

Product_Category	sale_volume
Electronics	381410.0
Grocery	357248.0
Clothing	293529.0
Books	292075.0
Home Decor	291014.0

Insight: Electronics has the highest sales volume.

---

Which product category generates the highest revenue?

product_revenue = df.groupby('Product_Category').agg(
    revenue = ('Total_Amount', 'sum'),
)
product_revenue = product_revenue.sort_values('revenue', ascending=False)
product_revenue

Out[16]:

Product_Category	revenue
Electronics	9.728525e+07
Grocery	9.096255e+07
Clothing	7.473945e+07
Books	7.454330e+07
Home Decor	7.417653e+07

Insight: Electronics has the highest revenue.

---

Which category has the highest AOV?

AOV (Average Order Value) — different definitions shown below.

Per-row AOV (Amount per purchase):

df['AOV'] = df['Amount'] / df['Total_Purchases']
df['AOV'].mean()

Out[17]:

np.float64(77.15064527254084)

Overall AOV (total revenue / total orders):

df['Total_Amount'].sum() / df['Total_Purchases'].sum()

Out[18]:

np.float64(254.88533467381427)

AOV by Product_Category:

pdf = df.groupby('Product_Category').agg(
    revenue = ('Total_Amount', 'sum'),
    orders = ('Total_Purchases', 'sum')
)
pdf['aov'] = pdf['revenue'] / pdf['orders']
pdf = pdf.sort_values('aov', ascending=False)
pdf

Out[19]:

Product_Category	revenue	orders	aov
Books	7.454330e+07	292075.0	255.219728
Electronics	9.728525e+07	381410.0	255.067380
Home Decor	7.417653e+07	291014.0	254.889899
Clothing	7.473945e+07	293529.0	254.623718
Grocery	9.096255e+07	357248.0	254.620185

Insight: Books have the highest average order value.

---

Which category receives the best feedback (Excellent %)?

Compute count of "Excellent" per category, total feedback per category, then percentage.

excellent_df = df.loc[df['Feedback'] == 'Excellent']
excellent_per_category = excellent_df.groupby('Product_Category').agg(
    excellent_count = ('Feedback', 'count')
)

total_df = df.groupby('Product_Category').agg(
    total_count = ('Feedback', 'count')
)

excellent_per_category['per'] = excellent_per_category['excellent_count'] / total_df['total_count'] * 100
excellent_per_category

Out[22]:

Product_Category	excellent_count	per
Books	19187	35.150038
Clothing	19183	35.065624
Electronics	24028	33.769483
Grocery	19268	28.863756
Home Decor	18980	34.926301

---

Which category has the highest and lowest average rating?

category_rating = df.groupby('Product_Category').agg(
    avg_rating = ('Ratings', 'mean'),
    median_rating = ('Ratings', 'median')
)
category_rating

Out[23]:

Product_Category	avg_rating	median_rating
Books	3.112428	3.0
Clothing	3.104266	3.0
Electronics	3.268857	4.0
Grocery	3.181739	3.0
Home Decor	3.109600	3.0

---

Which product category has the highest return on each customer?

highest return = revenue / unique customer

pdf = df.groupby('Product_Category').agg(
    revenue = ('Total_Amount', 'sum'),
    customer_count = ('Customer_ID', 'nunique')
)
pdf['highest return'] = pdf['revenue'] / pdf['customer_count']
pdf = pdf.sort_values('highest return', ascending=False)
pdf

Out[24]:

Product_Category	revenue	customer_count	highest return
Electronics	9.728525e+07	48945	1987.644285
Grocery	9.096255e+07	46823	1942.689529
Clothing	7.473945e+07	40770	1833.197086
Home Decor	7.417653e+07	40530	1830.163560
Books	7.454330e+07	40803	1826.907389

---

Which category has the highest ARPU?

ARPU = Revenue / Total Orders (orders counted as Total_Purchases sum)

adf = df.groupby('Product_Category').agg(
    revenue = ('Total_Amount', 'sum'),
    total_order = ('Total_Purchases', 'sum')
)
adf['aov'] = adf['revenue'] / adf['total_order']
adf

Out[27]:

Product_Category	revenue	total_order	aov
Books	7.454330e+07	54572	1365.962439
Clothing	7.473945e+07	54659	1367.376739
Electronics	9.728525e+07	71121	1367.883600
Grocery	9.096255e+07	66708	1363.592850
Home Decor	7.417653e+07	54306	1365.899332

---

Which category has the highest AOV? (alternate calculation)

AOV = Revenue / Number of Orders (counting orders as rows)

adf = df.groupby('Product_Category').agg(
    revenue = ('Total_Amount', 'sum'),
    total_order = ('Total_Purchases', 'count')
)
adf['aov'] = adf['revenue'] / adf['total_order']
adf

Out[28]:

Product_Category	revenue	total_order	aov
Books	7.454330e+07	54572	1365.962439
Clothing	7.473945e+07	54659	1367.376739
Electronics	9.728525e+07	71121	1367.883600
Grocery	9.096255e+07	66708	1363.592850
Home Decor	7.417653e+07	54306	1365.899332

---

(End of analysis)