Mall Customer Analytics — Strategy Dashboard

A beginner-to-intermediate data analysis project built with Python in a Jupyter Notebook. It loads mall customer data, cleans it, and produces a single multi-panel matplotlib/seaborn dashboard saved as final_mall_dashboard.png.

---

Table of Contents

1. Dataset Overview
2. Project Structure
3. What the Notebook Does — Step by Step
4. Dashboard Panels Explained
5. Key Insights from the Current Analysis
6. What Interns Can Improve
7. New Things to Practice for a Better Dashboard
8. How to Run

---

1. Dataset Overview

File: da_mall_customer.csv
Rows: 200 customers
Columns: 7

Column	Type	Description
CustomerID	Integer	Unique identifier for each customer
Gender	Categorical (M / F)	Customer gender
Age	Integer	Customer age (range ~18–70)
Education	Categorical	Highest education level — High School, Graduate, College, Post-Graduate, Doctorate, Uneducated, Unknown
Marital Status	Categorical	Married, Single, Divorced, Unknown
Annual Income (k$)	Integer	Annual income in thousands of USD
Spending Score (1-100)	Integer	Mall-assigned score based on customer spending behavior (1 = lowest, 100 = highest)

Notable data quality issues (handled in the notebook):

• The Education column had a trailing whitespace in its name ("Education ") — stripped during cleaning.
• Both Education and Marital Status contain "Unknown" entries that are relabeled to "Not Specified" for cleaner chart labels.

---

2. Project Structure

da_dashbrd/
├── anailyze.ipynb            # Main analysis notebook
├── da_mall_customer.csv      # Raw dataset (200 mall customers)
├── final_mall_dashboard.png  # Output dashboard image (auto-generated)
├── requirements.txt          # Python dependencies
└── README.md                 # This file

---

3. What the Notebook Does — Step by Step

Cell 1 — Import core libraries

import pandas as pd
import numpy as np

Loads pandas for data manipulation and numpy for numerical operations.

Cell 2 — Load and preview data

df = pd.read_csv('da_mall_customer.csv')
df.head()

Reads the CSV into a DataFrame and displays the first 5 rows to verify it loaded correctly.

Cell 3 — Import visualization libraries

import matplotlib.pyplot as plt
import seaborn as sns

Loads matplotlib (base charts) and seaborn (statistical visualizations on top of matplotlib).

Cell 4 — Clean column names

df.columns = df.columns.str.strip()

Removes any leading/trailing whitespace from all column headers. Without this, "Education " (with a trailing space) would cause KeyError failures later.

Cell 5 — Standardize unknown values

df['Education'] = df['Education'].replace('Unknown', 'Not Specified')
df['Marital Status'] = df['Marital Status'].replace('Unknown', 'Not Specified')

Replaces the literal string "Unknown" with "Not Specified" in both categorical columns for cleaner pie chart labels and better readability.

Cell 6 — Build and save the full dashboard

Creates an 18 × 12 inch figure with a 3 × 3 GridSpec layout containing 6 panels. The dashboard is saved as final_mall_dashboard.png and rendered inline.

---

4. Dashboard Panels Explained

Panel 1 — Business KPIs (Top Left)

A text-only panel displaying three headline numbers:

• Total Shoppers: 200
• Avg. Income: mean of Annual Income (k$)
• Avg. Spend Score: mean of Spending Score (1-100)

This gives an at-a-glance executive summary without any chart overhead.

Panel 2 — Customer Base by Gender (Top Center)

A seaborn.countplot showing how many Male vs Female customers are in the dataset. Quick check on demographic composition.

Panel 3 — Education Distribution (Top Right)

A matplotlib pie chart showing the percentage split across education levels (Graduate, High School, College, Post-Graduate, Doctorate, Uneducated, Not Specified).

Panel 4 — Age vs. Spending Score Trend (Middle Left + Center, spanning 2 columns)

A seaborn.regplot (scatter + regression line) plotting Age on the X-axis against Spending Score on the Y-axis. The red regression line shows the overall trend — generally, spending score decreases slightly as age increases.

Panel 5 — Income Distribution (Middle Right)

A seaborn.histplot with KDE (Kernel Density Estimate) overlay for Annual Income (k$). Reveals whether income is normally distributed, skewed, or bimodal.

Panel 6 — Market Segmentation: Income vs. Spending Score (Bottom Row, full width)

The most strategic panel. A seaborn.scatterplot mapping:

• X-axis: Annual Income (k$)
• Y-axis: Spending Score (1-100)
• Color (hue): Gender
• Point size: Age

Two dashed grey lines divide the space into quadrants:

• Vertical line at income = $60k
• Horizontal line at spending score = 50

Three quadrant labels are annotated:

Label	Quadrant	Meaning
IMPULSIVE	Low Income, High Spending	Spends a lot despite earning little
TARGET GROUP	High Income, High Spending	Best customers — high value, high engagement
HIGH POTENTIAL	High Income, Low Spending	Earns a lot but doesn't spend — opportunity for upselling

---

5. Key Insights from the Current Analysis

• The dataset is relatively balanced but slightly female-dominant.
• Graduate-level customers form the largest education segment.
• There is a mild negative correlation between age and spending score — younger shoppers tend to spend more.
• Income follows a roughly normal distribution with most customers earning $40k–$80k.
• The segmentation scatter reveals a clear cluster in the Target Group (high income + high spending) that warrants focused marketing.
• The High Potential segment (high income, low spending) is the biggest untapped opportunity.

---

6. What Can Improve

These are concrete improvements to the existing analysis:

Data Quality

• Handle the "Uneducated" and "Not Specified" labels more carefully — consider grouping them or excluding them from percentage calculations.
• Check for and handle duplicate CustomerIDs.
• Validate the Spending Score column for out-of-range values (should be 1–100).
• Add df.info() and df.describe() cells to document data types and basic statistics.

Chart Improvements

• Add value labels on top of the gender countplot bars (e.g., show exact counts).
• Add a fourth quadrant label ("SAVERS" — Low Income, Low Spending) to complete the segmentation story.
• Use consistent color palettes across all panels for a more professional look.
• Add axis labels and units to every chart (e.g., "Annual Income (USD thousands)").
• Increase font sizes for tick labels — currently hard to read at smaller display sizes.

Analysis Depth

• Add a Marital Status panel — it's in the dataset but not visualized at all.
• Break down Spending Score by Education using a boxplot to see which education group spends the most.
• Add a correlation heatmap (sns.heatmap) for numeric columns: Age, Income, Spending Score.
• Add a Gender × Income grouped bar chart to compare income levels between male and female customers.

---

7. New Things to Practice for a Better Dashboard

These are skills and tools interns can learn to significantly level up the dashboard:

Intermediate Python / Data Analysis

Skill	What to Practice
groupby + agg	Calculate average spending score per education level, per gender, per age group
pd.cut / pd.qcut	Create age buckets (18–25, 26–35, 36–50, 50+) and analyze each group separately
value_counts(normalize=True)	Convert counts to percentages for cleaner reporting
Boolean filtering	Isolate the "Target Group" customers and profile them separately

Machine Learning (next step after EDA)

Technique	Purpose
K-Means Clustering (sklearn.cluster.KMeans)	Automatically find customer segments from Income + Spending Score instead of using manual quadrant lines
Elbow Method	Determine the optimal number of clusters k
PCA (Principal Component Analysis)	Reduce dimensions for visualizing clusters when more features are added

Interactive Dashboards (replace static matplotlib)

Tool	Use Case
Plotly Express (plotly.express)	Drop-in replacement for seaborn — produces interactive hover charts
Dash by Plotly	Build a full web dashboard app in Python with dropdown filters and sliders
Streamlit	Fastest way to turn a notebook into a shareable web app — add a gender/income filter sidebar
Power BI / Tableau	Industry-standard BI tools — practice connecting CSV data and building the same dashboard visually

Code Quality & Reproducibility

Practice	Why It Matters
Write reusable functions for each chart panel	Keeps code DRY and easier to maintain
Add Markdown cells between code cells	Documents your reasoning — essential for professional notebooks
Pin library versions in requirements.txt (e.g., pandas==2.2.0)	Ensures the notebook runs identically on any machine
Use pathlib.Path instead of raw strings for file paths	Cross-platform compatibility (Windows vs Mac/Linux)

Storytelling & Presentation

• Structure the notebook as a story: Business Question → Data → Finding → Recommendation
• Add a final Markdown cell with a written summary of the 3 most important findings and their business implications
• Export the notebook as a PDF or HTML report (jupyter nbconvert) to share without requiring Python

---

8. How to Run

Prerequisites: Python 3.8+

# 1. Install dependencies
pip install -r requirements.txt

# 2. Open the notebook
jupyter notebook anailyze.ipynb
# or in VS Code: open anailyze.ipynb directly

# 3. Run all cells (Kernel > Restart & Run All)
# Output: final_mall_dashboard.png will be created in the same folder

---

Dataset: Mall Customer Segmentation — commonly used for beginner clustering and EDA practice.