RideIQ — Uber Product Analytics

Simulating a Senior Product Analyst role at Uber — analyzing 200,000+ rides across New York City to diagnose 3 critical business problems using SQL, Python, and Power BI.

---

The Problem

Uber's New York operations show a concerning pattern — 8 in 10 riders never return after their first month. Meanwhile, a significant portion of booked rides never complete, and surge pricing may be driving away demand rather than maximizing revenue.

This project investigates all three problems end-to-end — from raw data to business recommendations.

---

Business Problems

#	Problem	What We Found
1	Rider retention drop	D7: 34%, D30: 19% — 8 in 10 riders lost within a month
2	Cancellation funnel leakage	6 AM has lowest ride completion rate (81%)
3	Surge pricing impact	All hours fall in $10–$15 avg fare range (Medium Surge)

---

Key Findings

• Peak demand — Friday 7 PM peaks at 1,999 rides/hour
• Retention crisis — Only 19% of riders still active after 30 days
• Cancellation pattern — Early morning hours (6 AM) show highest cancellation rates
• Fare distribution — Average fare $11.36, median $8.50 — majority of rides are short trips
• Rider segments — 3 distinct segments identified: Low, Mid, and High value riders

---

Business Recommendations

Problem	Recommendation
Retention	Target D7 churned riders with discount voucher via push notification
Cancellation	Investigate early morning supply-demand gap — increase driver incentives at 6 AM
Surge	Run A/B test — cap surge at 1.5x in 2 cities for 30 days, measure demand vs revenue

---

Dashboard Preview

---

Tech Stack

Tool	Purpose
Python (Pandas, Matplotlib, Seaborn, Scikit-learn)	Cleaning, EDA, A/B testing, segmentation
SQL (PostgreSQL)	Funnel analysis, cohort retention, LTV, surge revenue
Power BI (DAX)	Executive dashboard — 3 pages, 5 DAX measures

---

Project Structure

rideiq-uber-analytics/
├── data/
│   ├── raw/                  ← Original Kaggle dataset
│   └── cleaned/              ← Processed dataset (195,065 rows)
├── notebooks/
│   ├── 01_data_cleaning.ipynb
│   ├── 02_eda.ipynb
│   ├── 03_ab_test.ipynb
│   └── 04_segmentation.ipynb
├── sql/
│   ├── 01_funnel.sql
│   ├── 02_retention_cohort.sql
│   ├── 03_ltv.sql
│   ├── 04_cancellation.sql
│   ├── 05_supply_demand.sql
│   └── 06_surge_revenue.sql
├── dashboard/
│   └── uber_dashboard.pbix
└── problem_statement.md

---

Notebooks

Notebook	Description
01_data_cleaning.ipynb	Cleaned 200K rows — removed 4,935 invalid records, fixed data types, engineered time features
02_eda.ipynb	8+ visualizations — hourly demand, fare distribution, day-of-week patterns, cohort heatmap
03_ab_test.ipynb	Simulated surge cap A/B test — Z-test, p-value: 0.46
04_segmentation.ipynb	KMeans clustering (k=3) — Low, Mid, High value rider segments

---

SQL Queries

File	Analysis
01_funnel.sql	Ride completion funnel by hour
02_retention_cohort.sql	Monthly cohort retention using window functions
03_ltv.sql	Rider lifetime value by passenger segment
04_cancellation.sql	Cancellation rate analysis by hour
05_supply_demand.sql	Peak demand hours by day of week
06_surge_revenue.sql	Revenue and surge level classification by hour

---

Dataset

Property	Value
Source	Uber Fares Dataset — Kaggle
Raw rows	200,000
Cleaned rows	195,065
Columns	11 (after feature engineering)
Location	New York City
Period	2009–2015

---

Built by Aditya Sharma · LinkedIn · GitHub