notebook
Peloton Customer Churn Analysis
December 2023 · DSCI 510 · Principles of Programming for Data Science
Investigating why Peloton customers reduce usage — combining 273 survey responses with geocoding (GeoPy) and U.S. Census demographics to surface churn patterns across age, income, and geography.
Built as the final project for DSCI 510 — Principles of Programming for Data Science at the USC Viterbi School of Engineering. The work extends an earlier Peloton essay (Pandemic Data Patterns) by moving from observational analysis to structured quantitative investigation.
What the project does
- Collects and cleans a 273-respondent customer survey on Peloton usage and churn.
- Geocodes respondent locations with the GeoPy API.
- Joins respondent geographies against U.S. Census demographic data (age, income, density).
- Visualizes churn patterns across demographic and geographic dimensions with Seaborn.
Why it matters
The ambition here was to stop treating churn as a single rate and start treating it as a distribution — one that varies meaningfully by life stage, income, and region. The notebook is structured so each analytical step (cleaning, joining, visualizing, interpreting) is separately reviewable — the point is the process as much as the findings.
Links
- Live project: viterbi-data-science-portfolio — Peloton
- Source: github.com/sorendeorlow/viterbi-data-science-portfolio
- Precursor essay: Leveraging multiple machine learning models to gain insights into evolving Peloton customer behavior (IDSN 590, 2022)