Overview

Welcome to my analysis of a superstore dataset, focusing on .... . This project was created out of a desire to empower , enhance and showcase my Python , Pandas , Matplotlib skills , it delves into the superstore dataset i got from Kaggle.

The data source is Kaggle Dataset Link which provides the foundation dataset for my analysis, containing information about a global superstore for 4 years such as customer names, segments, product categories and sub categories, country, sales... . Through a series of Python scripts, I explore key questions such as the most demanded product, salary trends...

The Questions

Below are the questions i want to answer in my project:

1. What is the most demanded product in the store ?
2. What region is the most profitable region for the store ?
3. How does the sales trend monthly for each year ?
4. Who are the top customers for the most profitable region ?
5. What are the top and lowest states in profit ?

Tools I Used

For my deep dive into the superstore dataset, I harnessed the power of several key tools:

• Python: The backbone of my analysis, allowing me to analyze the data and find the insights i wanted , also i used the following libraries throughout the project:
  • Pandas: This was used to analyze the data.
  • Matplotlib: Used this one to visualize the data.
  • numpy: Tiny usage for some operations.
  • calendar: Used for months analysis.
• Jupyter Notebooks: The tool i used to easily integrate my python scripts and run them.
• Visual Studio Code: My go-to for any project.
• Git & Github: Essential for version control and sharing my Python code and analysis.

Data Preparation and Cleanup

This section outlines the steps taken to prepate the data for analysis, ensurng accuracy and usability.

Import & Loading Data

I start by importing necessary libraries that i'll use throughout the project and then loading the dataset.

# Importing libraris ill use
import pandas as pd
import matplotlib.pyplot as plt
import numpy as np
import calendar
import humanize

# TODO : Loading the data
data = pd.read_csv('data/superstore.csv' , encoding='latin1')

Cleaning Data + Managing types

Because the dataset was already cleaned i added up some dummy data so i can apply some cleanings , below the scripts i wrote :

# TODO : Data Cleaning Phase + managing types
data['Order Date'] = pd.to_datetime(data['Order Date'])
data['Ship Date'] = pd.to_datetime(data['Ship Date'])

col_to_categorize = ['Ship Mode' , 'Sub-Category' , 'Category' , 'Region' , 'Segment' , 'State']
for col in col_to_categorize:
    data[col] = data[col].astype('category')

# Treat duplicate rows
duplicated_sum = data.duplicated().sum()
if duplicated_sum > 0 :
    data.drop_duplicates(inplace=True)

most_occurent_ship_mode = data['Ship Mode'].mode()[0]
most_occurent_segment = data['Segment'].mode()[0]

data['Ship Mode'] = data['Ship Mode'].fillna(most_occurent_ship_mode)
data['Segment'] = data['Segment'].fillna(most_occurent_segment)

# Check if any discounts higher then 1
discount_higher_then_01 = (data['Discount'] > 1).sum()
if discount_higher_then_01 >= 1:
    print("Discount Higher then 1 founded and fixed")
    data['Discount'] = np.where(data['Discount'] > 1 , data['Discount'].median() , data['Discount'])

Engineering features

Next step is to engineer features that i know i'll use throughout my project :

months = list(calendar.month_name)[1:]
data['Year'] = data['Order Date'].dt.strftime("%Y")
data['Month'] = data['Order Date'].dt.strftime("%B")

# Convert Month column to an ordered categorical type
data['Month'] = pd.Categorical(data['Month'] , categories=months , ordered=True)

data.set_index("Row ID" , inplace=True)

The Analysis

Each jupyter notebook for this project aimed at investigating specific aspect of the dataset to reply to the given question. Here how i approached each one :

1️⃣ What is the top demanded product in the store ?

To get the most demanded product in the store i grouped by the Product Name and counted the orders for that product , sorted then the result by the total orders count and finally extracted top 10 rows

View my notebook with detailed steps here : most_demanded_prooducts

Visualize Data

plt.figure(figsize=(14,6))
plt.barh(most_demanded_product.index, most_demanded_product.values)
plt.xlabel('Quantity Sold')
plt.title('Top 10 Most Demanded Products')
plt.gca().invert_yaxis()  # convert the largest on top
plt.savefig("figures/question_1_fig.png")
plt.show()

Insights

As we can see determine using the graph , Staple envelope is the most demanded product in the store with total of approximately 50 orders , Staples come in the second place with 45 orders , we can also realize that Staples Equipment in general are very demanded.

2️⃣ What region is the most profitable region for the store ?

To find the most profitable region in the dataset for the store first of all i grouped the data by Region focusing on Profit column , and got the sum of the profit per each region then i sorted the result by sum of the profit in descending order

View my notebook with detailed steps here : most_profitable_region

Visualize Data

plt.figure(figsize=(12,6))
most_profitable_regions_bars = plt.bar(most_profitable_regions.index , most_profitable_regions['Total Profit'])
plt.bar_label(most_profitable_regions_bars , label_type='edge' , padding=3)
plt.xticks(fontsize=6)
plt.xlabel("Region")
plt.ylabel("Total Profit in USD")
plt.title("What is the most profitable region for the store ?")
plt.show()

Insights

As we can see the most profitable region is The West Region collecting approximetaly $100k Total Profit over 4 years while the lowest is The Central Region with nearly $40k Total Profit.

3️⃣ How does the sales trend monthly for each year ?

To answer this question i decided to create a pivot table based on the data i have , with the index being the years , the columns are each month while the values will be the sum of the sales in that exact month for that year.

View my notebook with detailed steps here : sales_trend_monthly_year

Visualize Data

plt.style.use('ggplot')

plt.figure(figsize=(12 ,6))
for year in sales_by_month_yearly.index:
    plt.plot(months , sales_by_month_yearly.loc[year] , label=year , marker='o')
    max_month = sales_by_month_yearly.loc[year].idxmax()
    max_month_value = sales_by_month_yearly.loc[year].max()
    max_month_value_formatted = humanize.intword(max_month_value).replace(" thousand" ,  "k")
    month_index = months.index(max_month)
    plt.annotate(max_month_value_formatted , xy=(month_index , max_month_value) , xytext=(month_index - 0.4 , max_month_value - 1500)  , arrowprops=dict(arrowstyle='->',color='gray'))

plt.ylabel("Profit ($USD)")
plt.title("Total Profit Per Year Over The Months ($USD)")
plt.legend()
plt.savefig("figures/question_3_fig.png")
plt.show()

Insights

As we can see we can determine the the sales usually peaks in the last 4 months of the year September, October, November, December which makes sense, and also we can grasp that the highest month is November of 2017 with $118k in sales.

4️⃣ Who are the top customers for the most profitable region ?

To find the top customers in the most profitable region (West) , first of all i filtered the dataset to include the customers only from the most profitable region , grouped the result by customer ID and i also included his name and his state , i renamed the column to Total Profit , sorted the values in descending order by Total Profit and focused on the top 10 customers.

View my notebook with detailed steps here : top_10_profitable_customers

Visualize Data

plt.style.use('ggplot')
plt.figure(figsize=(12,6))
top_ten_prof_orders_bars = plt.bar(top_ten_profitable_customers.index , top_ten_profitable_customers['Total Profit'])
plt.bar_label(top_ten_prof_orders_bars , label_type='edge' , padding=3)
plt.xticks(fontsize=6)
plt.xlabel("Customer Name")
plt.ylabel("Total Profit in USD")
plt.title("Top 10 most profitable west customers ($USD)")
plt.savefig("figures/question_4_fig.png")
plt.show()

Insights

As we can the most profitable west customer is Raymond Buch with total profit for the score approximately $7k with significent difference of the second most profitable west customer Jane Waco with $2k total profit.

5️⃣ What are the top and lowest states in profit ?

To make this analysis first there is two parts Top States , Least States in Total Profit . First i grouped by the dataset by State and focused on Profit column to get the sum of Profit per state , then to get the too state i sorted by total profit in descending order and then extracted the top ten column in the other side for the least profitable states i filtered the dataset to get me only the rows where Profit isnt under 0 and then did the exact same steps.

View my notebook with detailed steps here : top_lowest_10_profitable_states

Visualize Data

# Top 10 states
plt.figure(figsize=(15 , 6))
profit_per_top_state_bars = plt.bar(top_ten_profitable_state.index , top_ten_profitable_state['Total Profit'])
plt.bar_label(profit_per_top_state_bars , label_type='edge' , padding=3)
plt.ylabel("Total Profit ($USD)")
plt.title("What are the 10 most profitable states ?")
plt.savefig("figures/question_5(1)_fig.png")
plt.show()

plt.figure(figsize=(15 , 6))
profit_per_least_state_bars = plt.bar(least_10_profitable_state.index , least_10_profitable_state['Total Profit'])
plt.bar_label(profit_per_least_state_bars , label_type='edge' , padding=3)
plt.ylabel("Total Profit ($USD)")
plt.title("What are the 10 least profitable states?")
plt.savefig("figures/question_5(2)_fig.png")
plt.show()

Insights

As we can determine that most profitable states are California , New York , Washington , Michigan with other states The West have two states while the East have three and South two.

While for the lowest states are the likes of Wyoming Despite being a West State , West Virginia , North and South Dakota , Maine.