by Aranya Rayed | Team Hephaestus
This repository contains the submission for The Manhattan Project’s Exploratory Data Analysis (EDA) Individual Challenge.
The mission was to dive deep into the Spotify dataset and extract actionable insights for Spotify Tracks, with a sharp focus on:
- Listener Behavior 👥🎧
- Track Performance 🎵🗓️
- Streaming Patterns 🎙️📈
I combined data cleaning, statistical exploration, and insightful visualization to uncover hidden trends and help Spotify make smarter, data-driven decisions about music trends, audience preferences, and content strategy.
Our analysis of the music catalog dataset uncovered the following highlights:
-
Listener-Friendly Mid-Ranges Dominate: Most tracks feature moderate energy and danceability, avoiding extreme highs or lows → this strategy ensures broad mass appeal and prevents listener fatigue.
-
Shift Toward Moodier Tones: Musical "positivity" (valence) has steadily declined over the decades, even as tracks remain energetic → this reflects a growing market for more emotionally complex or somber music.
-
Growing Multilingual Diversity: While English and Tamil historically dominate, languages like Hindi and Korean show significant growth → highlights the need for a globalized catalog and improved language metadata to serve a diverse audience.
-
Power Law in Popularity: A small number of tracks, artists, and nostalgic eras (the 70s & 2000s) drive the majority of listens → balance blockbuster promotion with discovery features for niche content to maximize engagement and retention.
- Clone this repository
- Install the required dependencies:
pip install pandas matplotlib seaborn scipy- README.md - This file.
- spotify_tracks.csv - The raw dataset used for the challenge.
- spotidy_data_description.csv - Supplementary dataset that describes every column in spotify_tracks.csv.
- Spotify_EDA.ipynb - The main Jupyter Notebook with all the code and analysis.
- Spotify Dataset EDA - A summary of my findings and recommendations.