Dataset Overview & Business Problem
Foundation of the analysis — understanding the data we're working with.
Business Problem
What audio and release characteristics drive streaming success on Spotify? This analysis helps labels, artists, and A&R teams understand the anatomy of a hit song — from danceability to release timing — enabling data-informed decisions on production, marketing, and release strategy.
Dataset Overview
The dataset contains 953 rows (952 valid after cleaning) and 24 columns including track metadata, release dates, platform presence (Spotify, Apple, Deezer, Shazam), and 7 audio features: danceability, valence, energy, acousticness, instrumentalness, liveness, and speechiness.
Data Quality
99.9%
1 row dropped (invalid streams)
Columns
24
Audio features + metadata
Top 20% Streams
57.2%
Pareto principle confirmed
Max Single Song
3.7B
Blinding Lights — The Weeknd
Year Coverage
93yrs
Songs from 1930 to 2023
Data Cleaning Note
Streams column was stored as string — converted to numeric. 1 row dropped with non-numeric value.
deezer_playlists and shazam_charts were also string columns, treated as supplementary info. No duplicates found. All audio features are integer percentages (0–100).General Performance Analysis
Top songs, artists, and stream distribution across the dataset.
Top 10 Most-Streamed Songs
Ranked by total Spotify streams. The Weeknd leads with Blinding Lights at 3.7B streams.
| # | Track | Artist | Year | Streams |
|---|
Pareto Distribution
What fraction of total streams comes from the top 20% of songs?
57.2%
of all streams generated
by the top 20% of songs
by the top 20% of songs
0%57.2% → top 20%100%
Key Finding
The streaming economy is highly concentrated. A hit-driven market where 190 songs account for majority of plays.
Top 10 Most-Streamed Artists
Total streams aggregated per primary artist. The Weeknd, Taylor Swift, and Ed Sheeran lead with 14B+ streams each.
Streams by Release Year (2010–2023)
2022 dominates in total streams volume, but older songs from 2017–2019 show the highest average streams per song — indicating sustained catalogue performance.
Insight — Newer vs Older Songs
Pre-2020 songs average 1.09B streams vs post-2020 songs averaging 342M streams. Older songs accumulate more streams over time — recency bias is offset by longevity. The most streamed songs were released 2015–2019.
Audio Feature Analysis
How audio characteristics correlate with streaming success.
Feature Correlation with Total Streams
Pearson correlation coefficients between each audio feature and streams. Negative values indicate slight inverse relationships.
Statistical Finding
No audio feature has a strong linear correlation with streams. The strongest signals are speechiness (−0.112) and danceability (−0.105) — both weakly negative. This confirms that virality depends more on timing, artist brand, and marketing than on audio signature alone.
Energy vs. Streams
Scatter plot of energy% against stream count. High energy songs don't consistently outperform lower-energy tracks.
BPM Distribution — Top 25% Songs
Tempo range breakdown among the highest-streaming tracks. The sweet spot is 100–120 BPM.
Happy vs. Sad Songs
Comparing average streams by valence (emotional tone). Contrary to expectation, sad songs perform slightly better.
Sad (valence <40)
506M avg
Happy (valence ≥60)
464M avg
Major key
535M avg
Minor key
486M avg
Finding
Sad/melancholic songs edge out happier ones by ~9%, and Major key songs perform 10% better than Minor. Emotional depth resonates with listeners.
Average BPM by Year (2010–2023)
Tempo trend over time. Modern music shows fairly stable tempo clustering around 118–126 BPM.
Artist & Collaboration Insights
Who dominates the charts, and does collaboration pay off?
Solo vs. Collaboration Performance
Average streams by artist count per track. Contrary to popular belief, solo songs outperform collaborations significantly.
568M
Solo Songs
(1 artist)
(1 artist)
+33% more streams
vs
428M
Collaborations
(2+ artists)
(2+ artists)
Insight
Solo tracks average 33% more streams than collaborations. This may reflect that already-popular solo artists drive the top charts, while collaborations are more common among mid-tier performers.
Average Streams by Musical Key
C# leads with the highest average streams per song among all keys, followed by E and D#.
Top 10 Artists by Total Streams
Aggregated streaming count across all songs in the dataset per primary artist.
| # | Artist | Total Streams | Visual Scale |
|---|
Time-Based Trends
How the music industry has evolved across release years.
Songs Released Per Year (2015–2023)
2022 had the most songs in the dataset (402), reflecting increased catalog volume and playlist placement competition.
Avg Streams Per Song by Release Year
Songs from 2016–2019 have the highest average per-song streams — they've had time to accumulate vs 2022–2023 songs which are still early in lifecycle.
Trend Finding — Streaming Surge Post-2020
Total streams volume grew dramatically from 2020 onwards. 2021 saw 73.8B total streams, and 2022 reached 116.4B — driven by the explosion of short-form content (TikTok virality) and playlist-driven discovery. Post-2020 represents the "streaming maturity era."
Advanced Business Insights
Predictive signals and the anatomy of a hit song.
What Defines a "Hit Song"?
Comparing audio feature averages: all songs vs. top 10% of streams.
Predictive Model Summary
Can we predict popularity from audio features alone?
Model · Linear Regression (Audio → Streams)
R² ≈ 0.03 — Audio features alone explain only ~3% of stream variance. Popularity is not predictable from sound alone.
Model · Playlist Inclusion → Streams
Spotify playlist count has the strongest positive correlation with streams. Being in 3,000+ playlists is the #1 predictor of high streams.
Combined Signals Model
Best hit formula: Release 2016–2019 · Major key · 100–130 BPM · Low speechiness · High playlist inclusion · Solo or power-duo artist.
Key Findings Summary
Distilled analytical conclusions from this report.
Pareto Principle Holds
Top 20% of songs (190 tracks) generate 57.2% of all streams. Streaming is a winner-takes-most market.
Audio Features ≠ Success
No audio feature has a correlation >|0.15| with streams. Sound design is not the primary driver of virality.
Catalogue Ages Like Wine
Songs from 2016–2019 average 3–4x the per-song streams of 2022 releases. Longevity multiplies performance.
Sad > Happy (Barely)
Low-valence songs average 506M streams vs 464M for high-valence tracks. Emotional weight resonates with listeners.
Solo Dominance
Single-artist tracks average 568M streams (+33% vs collabs). Top solo artists dominate the entire catalogue.
Sweet Spot: 100–120 BPM
The most common BPM range among top performers. Danceable but not frantic — optimized for playlist and club use.
Recommendations & Conclusion
Actionable insights for artists, labels, and A&R teams.
01
Prioritize Playlist Pitching
Playlist inclusion is the strongest predictor of high stream counts. Allocate marketing budget toward Spotify editorial pitching, playlist placement agencies, and independent playlist networks.
02
Invest in Back Catalogue
Pre-2020 songs massively outperform recent releases in per-song streams. Labels should run re-marketing campaigns on established catalogue — sync licensing, remastered versions, anniversary campaigns.
03
Target 100–130 BPM Range
Productions in the 100–130 BPM range appear most commonly among top performers. This tempo range suits both streaming and live performance — ideal for pop, hip-hop crossovers.
04
Emotional Depth Over Positivity
Low-valence songs slightly outperform. Lean into emotional storytelling, minor-to-major progressions, and authentic vulnerability in lyrics — not forced positivity.
05
Don't Over-Index on Audio Features
No single audio signature guarantees a hit. R² of ~0.03 means sound is 3% of the story. Focus on artist brand building, release timing, and cross-platform marketing — the other 97%.
06
Major Key + C# / E Tonality
Major key songs average 10% more streams than minor. C#, E, and D# average the highest per-song stream counts — worth considering during composition if targeting playlists.
Conclusion
This analysis reveals that the streaming economy is a hit-driven, catalogue-compounding market. Audio features alone are weak predictors of success — the real drivers are platform visibility (playlists), artist brand equity, and time on platform. A hit song in 2016 still outperforms a 2023 release in per-song streams. For stakeholders, the data-backed strategy is clear: build for longevity, optimize for playlist discovery, and don't chase audio trends — build emotional resonance. The Weeknd's dominance at 3.7B streams with Blinding Lights exemplifies all of this: a universally emotional song, released at peak Spotify adoption, in heavy playlist rotation for years.