🚖 Uber Ride Bookings Analysis — Python EDA

Tools used: Python | Pandas | NumPy | Matplotlib | Seaborn | Plotly
Status: Completed

---

📌 Short Description

This project performs an end-to-end Exploratory Data Analysis on an NCR (Delhi) Uber ride bookings dataset using Python. I cleaned the raw data, engineered new time-based features, and built 10+ visualizations to understand demand patterns, revenue drivers, vehicle preferences, booking status breakdown, and payment behavior. The goal was to find what is actually driving bookings and revenue — and where the business is losing money through cancellations.

---

🛠️ Tech Stack

Tool	Purpose
🐍 Python	Core language for the entire analysis
🐼 Pandas	Data loading, cleaning, feature engineering, and groupby analysis
🔢 NumPy	Numeric operations and coercion of messy columns
📊 Matplotlib	Static charts — line, bar, and heatmap visuals
🎨 Seaborn	Styled countplots and heatmaps
🎨 Seaborn	Histogram for ride distance distribution, scatter for distance vs booking value, heatmap for day vs hour demand
📓 Jupyter Notebook	Writing and running the full analysis interactively
📁 File Format	.ipynb for the notebook, .csv for the raw dataset

---

📁 Data Source

Source: NCR Ride Bookings Dataset (Kaggle)

The dataset contains individual ride booking records with the following key columns:

• Booking ID — Unique identifier per ride
• Date / Time — When the booking was made
• Vehicle Type — Auto, Go Mini, Go Sedan, Prime Sedan, Uber XL, etc.
• Booking Status — Completed, Cancelled by Driver, Cancelled by Customer, No Driver Found
• Booking Value — Fare amount in ₹
• Ride Distance — Trip distance in km
• Pickup Location / Drop Location — NCR locations
• Payment Method — UPI, Cash, Debit Card, Credit Card, Uber Wallet
• Driver Ratings / Customer Rating — Ratings out of 5

Features engineered during cleaning:

• Hour — Hour of booking extracted from DateTime
• DayofWeek — Day name from DateTime
• Month — Month number from DateTime
• Weekends — Boolean flag for Saturday and Sunday
• Is_Successful — True if booking status is Completed
• Is_Cancelled_Customer / Is_Cancelled_Driver — Cancellation flags
• Status Category — Simplified status: Completed / Cancelled / No Driver Found

---

🔍 Features / Highlights

🔴 Business Problem

An Uber-style ride booking platform in NCR has thousands of daily bookings — but a large chunk of them never complete. The business needed to understand:

• When is demand highest — and when does it crash?
• Which vehicle types are generating the most revenue?
• Why are 38% of bookings not completing?
• Who is cancelling more — drivers or customers?
• What payment methods do customers prefer?
• Does ride distance actually drive booking value?

---

🎯 Goal of the Analysis

To use Python to clean, engineer, and analyze NCR ride booking data — and find patterns in demand, revenue, cancellations, and customer behavior that the raw data alone cannot show.

---

📊 Walkthrough of Key Visuals

Executive KPIs (Printed Metrics)

Metric	Value
Total Rides Analyzed	1,50,000
Completed Rides	93,000
Completion Rate	62%
Total Revenue	₹4,72,60,574
Average Booking Value	₹508
Average Ride Distance	26.0 km
Avg Driver Rating	4.23 / 5
Avg Customer Rating	4.40 / 5
Customer Cancellations	10,500
Driver Cancellations	27,000
Peak Hour	6 PM
Peak Month	July

---

Demand Analysis — Bookings by Hour (Line Chart)

Ride demand follows a clear double-peak pattern —

• Morning peak around 10 AM — office commute hours
• Evening peak at 6 PM — the busiest hour of the entire day
• Lowest demand between 12 AM and 4 AM

---

Bookings by Day of Week (Bar Chart)

Demand is fairly consistent across all 7 days — no single weekday dominates. Weekends account for around 40% of all bookings despite being only 2 out of 7 days.

---

Demand Heatmap — Day vs Hour (Seaborn Heatmap)

Confirms that 5 PM to 7 PM is consistently the busiest time slot across every day of the week — not just weekdays. The heatmap makes this pattern immediately visible.

---

Monthly Bookings Trend (Line Chart)

• July is the peak month for bookings
• February is the lowest month
• No consistent upward or downward trend — demand fluctuates across the year

---

Revenue by Vehicle Type (Horizontal Bar Chart)

• Auto is the highest revenue-generating vehicle type at 24.8% of total revenue
• Uber XL contributes only 2% — lowest performing category
• Go Mini and Go Sedan are the next strongest performers

---

Revenue by Day (Bar Chart)

Weekends generate approximately 37% of total weekly revenue despite being only 2 days — a clear indicator of where demand concentration sits.

---

Revenue by Hour (Line Chart)

• 6 PM is the most profitable hour of the day
• Revenue peaks again around 10 AM
• Consistently low between 12 AM and 4 AM
• Sharp decline after 10 PM

---

Booking Status Breakdown (Pie Chart)

Status	Share
Completed	62%
Cancelled by Driver	18%
Cancelled by Customer	7%
No Driver Found	7%

38% of bookings are failing — and drivers are the biggest reason why.

---

Payment Method Distribution (Countplot)

• UPI dominates at ~45% of all transactions
• Debit Card, Credit Card, and Uber Wallet together make up ~30%
• Cash still accounts for ~24% of transactions

---

Ride Distance Distribution (Plotly Histogram)

Shows the spread of trip distances across all bookings — used to understand what the typical ride length looks like in NCR.

---

Ride Distance vs Booking Value (Plotly Scatter)

Scatter plot colored by vehicle type to check whether longer rides consistently produce higher fares — and whether vehicle type affects that relationship.

---

💡 Business Insights

• 6 PM is the most critical hour — both bookings and revenue peak here. Driver availability at this time directly impacts revenue.

• 38% of bookings are failing — this is a major operational problem. Drivers are responsible for 18% of all cancellations — nearly 3x the customer cancellation rate.

• Weekends punch above their weight — 2 days generating 37% of weekly revenue means weekend supply needs to match weekend demand.

• Auto, Go Mini, and Go Sedan carry the business — these three vehicle types dominate both bookings and revenue. Uber XL is barely contributing.

• UPI is already dominant at 45% — cash at 24% is still meaningful but digital is clearly winning.

• Demand is consistent across weekdays — no single weekday stands out. The real split is weekdays vs weekends, not Mon vs Fri.

• July is peak, February is lowest — operational planning and driver incentives should account for this seasonal swing.

---

📌 Recommendations

1. Fix the driver cancellation problem first — 18% driver cancellation rate is too high. Incentive structures or penalties need reviewing urgently.

2. Surge pricing at 6 PM — peak demand hour with consistent revenue. Pricing optimization here would directly improve margins.

3. Increase driver supply on weekends — 40% of bookings and 37% of revenue happen on 2 days. Driver shortages on weekends mean lost revenue.

4. Invest in Auto, Go Mini, Go Sedan — these three carry the business. Ensure high driver availability in these categories at all times.

5. Push UPI incentives further — already at 45% but reducing cash (24%) further would cut operational cost and improve settlement speed.

6. Run February promotions — lowest booking month. A targeted discount or referral campaign could lift demand during this slow period.

7. Address the No Driver Found problem — 7% of bookings failing because no driver was available means supply is not meeting demand in certain locations or hours.

📂 Project Structure

uber-ride-analysis/
├── Uber_ride_analysis.ipynb      ← Full analysis notebook
├── ncr_ride_bookings.csv         ← Raw dataset
└── README.md

---

💡 What I Learned

• Writing a reusable clean_data() function instead of cleaning inline — keeps the notebook organized
• Engineering time features from DateTime — Hour, DayofWeek, Month, Weekends flag
• Using pd.to_numeric(errors='coerce') to handle messy numeric columns without breaking the pipeline
• Choosing between Matplotlib and Seaborn depending on what the chart needs
• Using Seaborn for statistical charts like histogram, scatter, and heatmap where styling matters
• Thinking about cancellation rates as a business metric, not just a data cleaning step

---

🙋 Connect with Me

This is part of my ongoing data analytics learning journey. If you have feedback on the analysis, the code structure, or the charts — I would genuinely like to hear it.

• 🔗 LinkedIn: linkedin.com/in/tarun-kumar-5b9280396
• 💻 GitHub: github.com/tarunkumar7906

---

Still learning. Open to feedback.