Product Development at Uber Analysis

Table of Contents

• Project Overview
• Approach
• Key Findings
• Business Impact

Project Overview

High rider cancellations during peak commute hours were reducing retention and limiting profitability. We aimed to test whether adjusting wait times dynamically by time-of-day could improve rider satisfaction without increasing operational costs.

Approach

Summarization: Conducted A/B testing using Python (SciPy) to evaluate the impact of wait-time adjustments on rider behaviour, applying t-tests and hypothesis testing to assess statistical significance across time-of-day segments.

• Experimental Design

1. Built an A/B testing framework in Python (SciPy).

2. Split riders into control (current wait times) and treatment (adjusted wait times).

3. Applied t-tests and hypothesis testing to validate statistical significance.

• Segmentation

1. Evaluated results across time-of-day segments (morning/evening peak vs. off-peak).

2. Controlled for confounding variables such as route length and rider demographics to ensure robustness.

Key Findings

Summarization: Identified that shorter wait times during peak commute hours led to a significant drop in cancellations without raising operational costs, enabling the development of a dynamic, hour-specific strategy that improved rider retention and boosted profitability.

Peak Hours: Reducing wait times by 5 minutes led to a significant drop in cancellations (p < 0.05).

Off-Peak Hours: No material change in behaviour, suggesting that reducing wait times is unnecessary during low demand.

Operational Impact: Shorter wait times did not increase fleet or operating costs, making changes cost-neutral.

Business Impact

• Informed development of a dynamic, hour-specific scheduling strategy.

• Improved rider retention and boosted profitability for Uber.

• Provided a scalable, data-driven framework for future demand management.