🛵 Quick-Commerce Delivery Analytics

A complete end-to-end Data Analytics project on a real-world Quick-Commerce delivery dataset with 5,91,539 records — covering all 4 units of the Introduction to Data Analytics syllabus.

---

📌 Project Overview

This project performs a full data analytics study on a quick-commerce (q-commerce) delivery dataset simulating operations across multiple companies and cities in India. It covers everything from descriptive statistics to machine learning models, time series forecasting, and prescriptive decision-making.

Key Results

Metric	Value
Total Records Analysed	5,91,539
Columns	13
New Features Engineered	7
Hypothesis Tests Applied	5
ML Models Built	5+
Best Classifier (XGBoost) Accuracy	~79%
Best Classifier AUC-ROC	~0.88
K-Means Clusters Found	4
Best Forecast Model	Holt-Winters (MAE ~52)
Syllabus Units Covered	All 4

---

🗂️ Repository Structure

quick-commerce-analytics/
│
├── 📓 quick_commerce_analytics.ipynb     ← Main analysis notebook
├── 📊 Requirements.txt                   ← Python dependencies
├── 📄 README.md                          ← This file
├── 📄 LICENSE                            ← MIT License
├── 📄 Contributing.md                    ← Contribution guidelines
├── 📄 .gitignore                         ← Git ignore rules

Dataset: Download from Kaggle and place as quick_commerce_dataset.csv in the root folder.

---

📊 Dataset

Property	Details
Name	Quick-Commerce Delivery Dataset
Source	Kaggle — Rohit Grewal
Records	5,91,539 rows
Columns	13
Domain	Q-Commerce / On-Demand Delivery (India)
Access	Free on Kaggle

Column Descriptions

Column	Type	Description
Order_ID	Categorical	Unique identifier for each order
Company	Categorical	Q-commerce company (Blinkit, Zepto, Instamart, etc.)
City	Categorical	City where the order was placed
Customer_Age	Numerical	Age of the customer (years)
Order_Value	Numerical	Total monetary value of the order (₹)
Delivery_Time_Min	Numerical	Time taken to deliver (minutes)
Distance_Km	Numerical	Distance from dark store to customer (km)
Items_Count	Numerical	Number of items in the order
Product_Category	Categorical	Primary product category
Payment_Method	Categorical	Mode of payment used
Customer_Rating	Numerical	Customer satisfaction rating (1–5 stars)
Discount_Applied	Categorical	Whether a discount was used (Yes/No)
Delivery_Partner_Rating	Numerical	Rating given to delivery partner (1–5 stars)

---

🤖 Analytics & ML Techniques

#	Technique	Type	Purpose
1	Descriptive Statistics	Statistical	Mean, Median, Std, Skewness, IQR, Percentiles
2	Exploratory Data Analysis	Visual	Histograms, Box plots, Scatter, Pair plots
3	Feature Engineering	Data Prep	7 new features created from raw columns
4	Central Limit Theorem	Probability	Sampling distribution demonstration
5	Confidence Intervals	Inferential Stats	90%, 95%, 99% CI for mean order value
6	t-test (One & Two-sample)	Hypothesis Testing	Discount effect on order value
7	One-Way ANOVA	Hypothesis Testing	Order value differences across companies
8	Two-Way ANOVA	Hypothesis Testing	Company × Product_Category interaction
9	Chi-Square Test	Hypothesis Testing	Payment method vs discount association
10	Pearson & Spearman Correlation	Statistical	Feature relationship analysis
11	Simple Linear Regression	Predictive	Items_Count → Order_Value
12	Multiple Linear Regression	Predictive	Multi-feature → Order_Value + What-If
13	Ridge & Lasso Regression	Regularised ML	Overfitting control
14	Decision Tree Classifier	ML Classification	High-value order prediction
15	Random Forest Classifier	ML Classification	Ensemble high-value prediction
16	XGBoost Classifier	ML Classification	Best-performing classifier
17	K-Means Clustering	Unsupervised ML	4-segment customer segmentation
18	Moving Averages (SMA, WMA)	Time Series	Trend smoothing
19	Exponential Smoothing (Holt-Winters)	Time Series	Trend + seasonality forecasting
20	Autoregressive AR(7) Model	Time Series	Lag-based demand forecasting
21	Linear Programming	Prescriptive	Delivery partner zone allocation
22	Decision Theory (EMV, Maximax, etc.)	Prescriptive	Optimal discount strategy selection

---

👥 Customer Segments Discovered (K-Means, K=4)

Cluster 0 — Premium Bulk Buyers (High Value, High Items)

• Average order value ~₹900 with ~9 items per order
• Longest delivery distance (~7 km) — outer zones
• Strategy: Offer loyalty subscriptions and priority delivery

Cluster 1 — High-Spend, Low-Item Buyers (Premium Category)

• Average order value ~₹750 with only ~3 items
• Higher ratings (~3.8) — satisfied premium customers
• Strategy: Promote premium single-item categories and branded products

Cluster 2 — Satisfied Budget Shoppers

• Average order value ~₹300 with ~5 items
• Highest customer rating (~4.2) — happiest segment
• Strategy: Discount bundles, combo offers, and repeat-order nudges

Cluster 3 — Quick Low-Value Shoppers

• Average order value ~₹180 with ~2 items
• Lowest rating (~2.8) — at-risk for churn
• Strategy: Minimum order incentives, free delivery thresholds

---

🔄 Project Pipeline

Raw Dataset (5,91,539 rows × 13 cols)
        ↓
Data Cleaning & Validation
  ├── Missing value check (0 nulls found)
  ├── Duplicate removal
  ├── IQR outlier detection
  └── Categorical encoding
        ↓
Feature Engineering (7 new features)
  ├── Value_Per_Item
  ├── Delivery_Speed
  ├── High_Value_Order (classification target)
  ├── Age_Group
  ├── Order_Efficiency
  ├── Rating_Gap
  └── Log_Order_Value
        ↓
Descriptive Statistics + EDA
        ↓
Hypothesis Testing (5 tests)
  ├── One-Sample t-test
  ├── Independent t-test (Welch's)
  ├── One-Way ANOVA + Tukey HSD
  ├── Two-Way ANOVA
  └── Chi-Square Test
        ↓
┌─────────────────────────────────────────┐
│  Regression Models                      │
│  ├── Simple Linear Regression           │
│  ├── Multiple Linear Regression         │
│  ├── Ridge Regression (α=1.0, 10.0)     │
│  └── Lasso Regression (α=0.1, 1.0)      │
└─────────────────────────────────────────┘
        ↓
┌─────────────────────────────────────────┐
│  Classification Models                  │
│  ├── Decision Tree  (Acc ~75%, AUC 0.83)│
│  ├── Random Forest  (Acc ~78%, AUC 0.87)│
│  └── XGBoost        (Acc ~79%, AUC 0.88)│
└─────────────────────────────────────────┘
        ↓
K-Means Clustering (K=4 optimal)
  ├── Elbow Method
  ├── Silhouette Score
  └── Calinski-Harabasz Score
        ↓
Time Series Forecasting
  ├── Moving Averages (SMA-7, SMA-30, WMA-7)
  ├── ADF Stationarity Test
  ├── ACF / PACF Analysis
  ├── Seasonal Decomposition
  ├── SES → Holt's → Holt-Winters ← BEST
  └── AR(7) Autoregressive Model
        ↓
Prescriptive Analytics
  ├── Linear Programming (partner allocation)
  └── Decision Theory (EMV, Maximax, Maximin, etc.)
        ↓
Business Insights & Recommendations

---

🗒️ Notebook Sections

Section	Description
1	Project Introduction & Industry Overview
2	Environment Setup & Library Installation
3	Data Loading & Structural Audit
4	Data Cleaning & Preprocessing
5	Feature Engineering (7 features)
6	Descriptive Statistics
7	Exploratory Data Analysis (EDA)
8	Delivery Performance Analysis
9	Customer Experience Analysis
10	Correlation & Covariance Analysis
11	Probability Theory & Hypothesis Testing
12	Regression Analysis + What-If Scenarios
13	Decision Tree Classification
14	Random Forest & XGBoost Classification
15	K-Means Clustering
16	Time Series Forecasting
17	Prescriptive Analytics
18	Business Insights
19	Business Recommendations
20	Conclusion & Summary

---

⚙️ Setup & Usage

Prerequisites

• Python 3.10+
• Google Account (for Colab)
• Kaggle account (to download dataset)

Step 1 — Open in Google Colab

Go to colab.research.google.com, click File → Upload Notebook, and select quick_commerce_analytics.ipynb.

Step 2 — Download the Dataset

Download from Kaggle and upload the CSV to your Colab session:

from google.colab import files
uploaded = files.upload()

Or mount Google Drive and load:

from google.colab import drive
drive.mount('/content/drive')
df = pd.read_csv('/content/drive/MyDrive/quick_commerce_dataset.csv')

Step 3 — Install Dependencies

The first notebook cell handles installation:

pip install xgboost statsmodels

All other libraries (pandas, numpy, scikit-learn, matplotlib, seaborn, scipy) are pre-installed on Colab.

Step 4 — Run All

Runtime → Run All (Ctrl+F9)

---

📦 Dependencies

# Install in Colab (only these 2 needed — rest are pre-installed)
xgboost>=1.7.0
statsmodels>=0.14.0

# Pre-installed on Google Colab
pandas>=1.5.0
numpy>=1.23.0
matplotlib>=3.6.0
seaborn>=0.12.0
scikit-learn>=1.2.0
scipy>=1.10.0

---

📈 Key Findings

Hypothesis Testing Results

Test 1 — One-Sample t-test        : p < 0.05 → Mean order value ≠ ₹500 ✗ H₀ rejected
Test 2 — Independent t-test       : p < 0.05 → Discounts increase order value ✗ H₀ rejected
Test 3 — One-Way ANOVA            : p < 0.05 → Companies differ in order value ✗ H₀ rejected
Test 4 — Chi-Square Test          : p < 0.05 → Payment ↔ Discount associated ✗ H₀ rejected
Test 5 — Two-Way ANOVA            : p < 0.001 → Company × Category interaction ✗ H₀ rejected

Top Predictors of Order Value

1. Items_Count        (r = +0.82) ← Strongest predictor
2. Value_Per_Item     (r = +0.75) ← Basket quality signal
3. Delivery_Speed     (r = −0.55) ← Operational proxy
4. Distance_Km        (r = +0.40) ← Moderate positive
5. Customer_Rating    (r = +0.02) ← Near-zero impact

SLA Compliance

Delivered within 30 min : ~50%  ← SLA target
Delivered within 45 min : ~75%
Delivered within 60 min : ~95%
Average delivery time   : ~30 min

Classification Model Comparison

Decision Tree  → Accuracy: ~75%  |  AUC: ~0.83
Random Forest  → Accuracy: ~78%  |  AUC: ~0.87
XGBoost        → Accuracy: ~79%  |  AUC: ~0.88  ← BEST

---

🎯 Business Recommendations

Area	Priority	Key Action
Inventory Planning	🟢 HIGH	Pre-stock top SKUs Fri–Sat; bundle promotions for 8+ item orders
Delivery Optimisation	🟢 HIGH	Open dark stores in SLA-breach zones; target <3 km radius
Pricing & Discounts	🔵 MEDIUM	Deploy 5% discount for orders ≥5 items (EMV optimal strategy)
Customer Experience	🔵 MEDIUM	Set 25-min internal SLA; invest in partner training
Marketing Strategy	⚪ ONGOING	Design campaigns per cluster; use ML for priority queuing

---

👨‍💻 Author

Aditya Manoj

---

📄 License

This project is licensed under the MIT License — see the LICENSE file for details.

---

🙏 Acknowledgements

• Rohit Grewal for the Quick-Commerce dataset on Kaggle
• Scikit-learn for ML algorithms
• Statsmodels for statistical modelling and time series
• XGBoost for gradient boosting classification
• Google Colab for free GPU/TPU compute

---

Built with ❤️ using Google Colab + Python + Scikit-learn