Amazon Search Bias: A Data-Driven Look

Project Overview

This project explores Amazon’s self-preferencing practices — how the platform may rank its own products, such as Amazon Basics, higher in search results than third-party competitors. Using a dataset collected via browser extensions from a panel of real users, the analysis investigates the spatial placement of products, the influence of branding, and the association between Prime eligibility and targeted brand promotion. This study is rooted in concerns around algorithmic bias and market fairness, especially in powerful e-commerce platforms like Amazon.

View Full Report

Key Findings

• Amazon ranks its own products more prominently than major brand or third-party items, especially in premium search result positions.
• There is a strong correlation between spatial placement (top-left of screen) and likelihood of being an Amazon-owned product — with a Pearson coefficient of 0.85.
• Products associated with Amazon Prime and targeted brands are significantly more likely to appear higher in search results, as confirmed by a chi-square test (p < 0.001).
• Amazon-branded products are cheaper, receive more reviews, and have higher average ratings than third-party items — possibly skewing perceived product value.
• While correlation doesn’t confirm causation, the pattern of recurrent preferential placements across multiple variables suggests algorithmic favoritism.

Jupyter Notebook Analysis

Below is an interactive view of the Jupyter Notebook analysis for this project.

Download Report

Click the button below to download the full pdf.