Understanding the Elements of Spotify Song Success

Author

Perry Lampertius

Introduction

Music plays a huge role in culture and daily life, and with platforms like Spotify and Apple Music, it’s more accessible than ever. Millions of people stream music every day, but not every song gets the same amount of attention. Some tracks go viral and hit the top of the charts, while others stay under the radar. This made me wonder—what actually makes a song successful on a streaming platform like Spotify? Is it the tempo, how danceable it is, how energetic it sounds, or maybe even what day it was released? Do these kinds of musical and platform-related features really affect how popular a song becomes?

As someone who enjoys music and regularly follows what’s trending on charts and playlists, I wanted to look into which features make a song more likely to become a hit. I think understanding these patterns could be helpful not only for music fans but also for artists, producers, and marketers who want to know what drives a song’s success in today’s streaming world.

To explore this, I chose a dataset that tracks the performance and characteristics of songs across major streaming platforms, with a focus on Spotify. This dataset includes information like stream counts, playlist and chart appearances, and audio features such as tempo, energy, and mood. My goal is to analyze these variables and see which ones are most strongly connected to popularity. Can we find any trends or clues that explain why some songs take off while others don’t? That’s what I’ll be exploring in this project.

Dataset Overview

This dataset, (click here to download), offers a detailed snapshot of what drives song popularity across major music streaming platforms, including Spotify, Apple Music, Deezer, and Shazam. It brings together both quantitative performance metrics—such as total streams, playlist appearances, and chart positions—and a wide range of musical features like tempo (BPM), energy, danceability, valence (mood), and more technical aspects such as key, mode, and acousticness. Additionally, it includes contextual information like release dates and artist count, allowing for analysis over time or by artist collaboration scale.

By combining musical characteristics with platform-specific performance data, the dataset provides a unique opportunity to explore not only which types of songs become popular, but also why they might succeed. For example, are high-energy songs more likely to be featured in popular playlists? Do slower tempos dominate certain seasons or genres? Does the number of artists on a track correlate with higher chart performance?

This dataset can be used to examine trends in music consumption, identify the traits of viral or commercially successful tracks, and even inform recommendations for artists and producers looking to optimize their sound for streaming platforms. It’s also valuable for academic research, marketing insights, and anyone interested in the intersection of music, data science, and digital media trends. In an era where streaming data plays a significant role in shaping the music industry, datasets like this one offer a window into the mechanisms of musical success in the digital age.

Key Variables in the Dataset

Basic Track Info

track_name: Title of the song
artist(s)_name: Name(s) of the performer(s)
artist_count: Number of credited artists
released_year, released_month, released_day: Release date (year, month, day)

Platform Performance

streams: Total number of streams (Spotify)
in_spotify_playlists, in_apple_playlists, in_deezer_playlists: Playlist count by platform
in_spotify_charts, in_apple_charts, in_deezer_charts, in_shazam_charts: Chart appearances per platform

Musical Features

bpm: Tempo (beats per minute)
key: Musical key (e.g., C, D#)
mode: major key and minor key

Audio Characteristics

danceability_%: Danceability score
valence_%: Positivity or happiness of the song
energy_%: Intensity level
acousticness_%: Degree of acoustic sound
instrumentalness_%: Likelihood of having no vocals
liveness_%: Likelihood the song was performed live
speechiness_%: Amount of spoken words

Summary Table of Statistics

Summary Table of Song Statistics
Variable	Statistic	Value
streams	Mean	12,112,503,461.40
streams	Min	2,762.00
streams	Max	11,053,756,970,173.00
streams	SD	358,050,104,389.57
bpm	Mean	122.54
bpm	Min	65.00
bpm	Max	206.00
bpm	SD	28.06
danceability_%	Mean	66.97
danceability_%	Min	23.00
danceability_%	Max	96.00
danceability_%	SD	14.63
valence_%	Mean	51.43
valence_%	Min	4.00
valence_%	Max	97.00
valence_%	SD	23.48
energy_%	Mean	64.28
energy_%	Min	9.00
energy_%	Max	97.00
energy_%	SD	16.55
acousticness_%	Mean	27.06
acousticness_%	Min	0.00
acousticness_%	Max	97.00
acousticness_%	SD	26.00
instrumentalness_%	Mean	1.58
instrumentalness_%	Min	0.00
instrumentalness_%	Max	91.00
instrumentalness_%	SD	8.41
liveness_%	Mean	18.21
liveness_%	Min	3.00
liveness_%	Max	97.00
liveness_%	SD	13.71
speechiness_%	Mean	10.13
speechiness_%	Min	2.00
speechiness_%	Max	64.00
speechiness_%	SD	9.91

The dataset features a wide mix of music, covering a broad spectrum of styles, tempos, and moods. Some tracks are fast-paced and high-energy, clearly designed for dancing or workout playlists, while others lean more acoustic and relaxed, likely intended for quieter, more introspective listening. This musical diversity makes the dataset rich for analysis, allowing comparisons across different genres and emotional tones.

One of the most striking aspects is the dramatic difference in stream counts. While some songs have only a few thousand streams, others have reached into the billions. This shows that music streaming success is far from evenly distributed—a small group of massively popular songs captures the vast majority of listener attention. The high standard deviation in stream counts further confirms this imbalance, highlighting how just a few viral or globally successful tracks can heavily skew the data. These outliers are likely boosted by factors like playlist placements, artist fame, and algorithmic promotion.

Looking at the musical features, there’s significant variation in metrics like BPM (tempo), energy percentage, and danceability. This opens the door for deeper exploration—such as clustering songs by genre, mood, or even by their likelihood of appearing on curated playlists. Despite the overall diversity, a common pattern emerges: most songs in the dataset are vocal-heavy and high in energy, while purely instrumental tracks are quite rare. This suggests that the dataset is tilted toward modern, mainstream pop music that’s crafted to be emotionally engaging and instantly catchy—music made to grab attention quickly in a fast-moving digital landscape.

In short, the dataset offers a vivid snapshot of today’s popular music culture: bold, loud, energetic, and emotionally charged, with a strong commercial focus on what resonates with large audiences on streaming platforms.

Visualizing the Trends Behind Spotify’s Most Popular Songs

Distribution of Spotify Streams

This chart gives us a good look at how popular songs are on Spotify. Most songs don’t get a ton of streams, but a few really famous ones get massive numbers. That makes it hard to compare them on a regular chart, so using a log scale on the x-axis helps a lot. It basically evens things out so we can actually see the differences without the super popular songs completely overshadowing the rest.

With the log scale, we can spot more patterns—like the fact that a lot of songs seem to land around the 100 million stream mark. That seems to be a sweet spot for songs that are doing really well, even if they’re not global hits. It shows there’s a big gap between your average song and the ones that completely blow up. Overall, this chart helps break down those differences and makes it easier to understand what success looks like on Spotify.

Relationship Between Energy and Stream Count

Looking at the data, there seems to be a bit of a connection between how energetic a song is and how many streams it gets, but it’s not super strong. A lot of the songs with around 75% energy do seem to have higher stream counts, so that could mean people are drawn to tracks that feel more upbeat or intense. Still, the trend doesn’t hold up across the board. There are plenty of high-energy songs that didn’t take off, and some chill, low-energy ones that ended up doing really well.

So yeah, energy might help a song stand out or make it more playlist-friendly, but it’s definitely not the only thing that matters. Streaming success probably comes down to a bunch of different factors—like who the artist is, how the song gets promoted, whether it lands on a big playlist, and even timing. In the end, energy alone doesn’t explain why a song blows up, but it might give it a bit of an edge.

Density of Streams by BPM

The peak in the graph indicates that songs within the 110–115 BPM range tend to receive higher stream counts overall, suggesting that this tempo is especially popular among listeners. This aligns with the fact that many mainstream pop and dance tracks fall within this range—fast enough to be energetic, but still easy to follow and accessible. However, it’s important to note that not every song in this tempo range is a hit. There are plenty of tracks with similar BPMs that haven’t gained much traction, and many successful songs fall outside of this range as well.

Flatter areas of the graph show that songs at certain other tempos typically receive fewer streams on average, even if there are a large number of them. Because the density plot is weighted by stream count, just a few highly popular songs can create a noticeable peak, even if that tempo isn’t the most common overall.

While this trend suggests that certain tempos may be more appealing to broad audiences, BPM alone doesn’t determine a song’s success. A track’s popularity is likely influenced by a mix of factors, including artist recognition, marketing, playlist placement, production quality, and cultural trends. Tempo may contribute to a song’s appeal, but it’s just one piece of a much larger picture.

Stream Popularity by Mode (Major/Minor)

This bar graph suggests that songs in a major key tend to receive more average streams than those in a minor key. In music theory, major keys are often associated with happier, brighter, and more uplifting emotional tones, while minor keys are typically perceived as more melancholic, introspective, or serious.

From a commercial perspective, audiences often gravitate toward songs that evoke positive emotions, especially in mainstream pop, dance, and commercial radio formats. Major key songs may therefore enjoy broader appeal, making them more likely to be added to popular playlists, shared by listeners, or used in upbeat media and advertisements. These factors can lead to a positive feedback loop, where major key songs get more exposure and, as a result, accumulate more streams.

Additionally, music production and songwriting trends often favor the major key for hit songs, partly because of its emotional impact and partly due to market-tested formulas. Artists and producers aiming for viral success might deliberately choose major keys for this reason.

However, it’s important to acknowledge that musical taste is diverse, and many successful songs are written in minor keys—especially in genres like hip-hop, alternative, R&B, or classical-influenced pop. Still, the aggregate data suggests a general preference or market trend favoring the emotional brightness of major-key songs in the streaming landscape.

Correlation between Speechiness and Spotify Streams

The analysis indicates that songs exhibiting moderate levels of speechiness—specifically in the range of 4–5%—tend to receive higher streaming counts compared to tracks with either minimal or excessive speech-like characteristics. This trend suggests a potential relationship between the incorporation of spoken-word elements and a song’s commercial appeal on platforms such as Spotify.

In the context of audio feature analysis, speechiness refers to the extent to which a track contains spoken words. Lower values generally represent purely sung or instrumental compositions, while higher values are characteristic of tracks dominated by speech, such as spoken word recordings, skits, or a cappella rap.

The observed peak in streaming performance at moderate speechiness levels implies that audiences may gravitate toward songs that balance traditional musical structure with subtle spoken or rhythmic vocal components. These may include: Conversational or narrative-style hooks, Rap verses integrated within melodic pop tracks, Spoken intros, ad-libs, or rhythmic vocal interjections, Such stylistic choices are particularly common in mainstream genres like pop, hip-hop, reggaeton, trap, and electronic dance music, where the fusion of melodic and rhythmic vocal delivery enhances listener engagement and emotional impact.

This pattern may reflect consumer preferences for music that is energetic, dynamic, and relatable, as speech-like inflections often convey attitude, authenticity, or immediacy. Conversely, tracks with very high speechiness may lack the melodic elements that typically drive streaming popularity, limiting their resonance with broader audiences.

Average Streams by Song Type (Solo vs Collaboration)

The analysis shows that, on average, solo songs receive more streams than collaborative tracks, challenging the belief that collaborations automatically result in higher streaming numbers due to the combined fan bases.

Several factors may explain this. Artist Popularity Bias is one key element. Influential solo artists, such as Taylor Swift or Drake, often have a dedicated following that drives their streaming success, regardless of collaborations. Their star power can overshadow any potential boost a collaboration might offer.

Another factor is Diluted Audience Appeal. While collaborations aim to unite fan bases, they don’t always align in terms of audience tastes. Combining different styles or genres can lead to a product that lacks cohesion, potentially alienating listeners. Additionally, collaborations may not capture the same authenticity or creative focus that solo artists can achieve on their own.

Marketing Power and Brand Identity also play a role. Solo artists have clear personal brands that resonate with their fans, and their promotions tend to be more focused. In contrast, collaborations often require joint marketing strategies, which can dilute the message and reduce impact.

While collaborations can be strategic for expanding reach or exploring new ideas, this data suggests they don’t necessarily lead to greater streaming success.

Average Streams by Month of Release

The chart indicates that January exhibits the highest average number of streams among all months, suggesting that songs released at the beginning of the year tend to achieve stronger streaming performance on average.

Several factors may contribute to this pattern. One potential explanation is seasonal listening behavior. Following the holiday season, many individuals resume regular routines—such as commuting, exercising, or working—which are often accompanied by increased music consumption. January may also present a less saturated release environment, as it typically sees fewer high-profile music releases compared to more competitive months like November and December. This relative lack of competition may provide songs released in January with greater visibility and a longer lifespan on playlists and charts.

In addition, the start of a new year often prompts both streaming platforms and listeners to refresh playlists and explore new content, which may further enhance the exposure and streaming performance of songs released during this period.

Comparison of Streaming Data with Billboard Top 100 Rankings

After analyzing the dataset and identifying key variables that influence a song’s popularity based on stream count, I would like to validate whether the current Top 100 Billboard rankings on Spotify align with the insights from the graphs created above. To do this, I will scrape data from the website https://kworb.net/spotify/, specifically collecting the daily Top 5 rankings from several countries over a two-week period. This timeframe will provide sufficient data to analyze which artists appear most frequently and how their popularity varies across different countries.

I have chosen the following countries for analysis: United States, United Kingdom, Brazil, Germany, and France. These countries were selected because they represent large Spotify audiences with diverse musical tastes, providing a broad perspective on global streaming trends.

Artist Distribution in Top 10

# A tibble: 7 × 2
  Artist                                       appearances
  <chr>                                              <int>
1 Alex Warren - Ordinary                                 5
2 Benson Boone - Beautiful Things                        5
3 Billie Eilish - BIRDS OF A FEATHER                     5
4 Lady Gaga - Abracadabra                                5
5 Lady Gaga - Die With A Smile (w/ Bruno Mars)           5
6 ROSÉ - APT. (w/ Bruno Mars)                            5
7 Teddy Swims - Lose Control                             5

The table shows 11 different songs and artists, and each one shows up exactly five times in the dataset. That means every song was popular in five different countries, but no one song or artist appeared more than the others. Even artists like Billie Eilish and Lady Gaga, who had more than one song on the list, still only showed up five times each. So, no one really stood out as being the most popular overall. This tells me that music tastes were pretty evenly spread out, and people in different countries were listening to a good mix of songs—not just a few big global hits.

Total Streams for Top 5 Artists

The chart shows the total streams for the top five artists, and Jennie’s song “Like Jennie” is way ahead of the rest—almost three times more streams than the second artist. That kind of lead really makes her stand out in the dataset. The other artists still did well, but their numbers are a lot lower, which shows that Jennie’s song got way more attention overall.

This big gap could mean Jennie’s track had a stronger global reach—maybe it got more playlist placements, more promotion, or just really clicked with people across different countries. Whatever the reason, “Like Jennie” clearly comes out on top and shows a level of popularity that the others didn’t quite match.

Conclusion

After going through all the data, one thing is super clear—there’s no exact recipe for what makes a song go viral on Spotify. Features like energy, BPM, or even the month it drops definitely matter, but they’re not a guarantee. A track can check all the boxes and still flop, while another song that breaks all the “rules” might end up everywhere. That tells me it’s not just about how a song sounds—it’s also about timing, marketing, who the artist is, and maybe even just being in the right place at the right time.

One of the biggest things that stood out to me was how uneven streaming success is. A few songs rake in billions of streams, while most barely scratch the surface. Jennie’s “Like Jennie” was a perfect example of this—it totally dominated the top 5 and showed how one song can take off way more than the rest. But on the flip side, there were also artists like Billie Eilish and Lady Gaga whose songs were more evenly spread out across countries, showing that music taste can be really diverse depending on where you are.

At the end of the day, the data shows that streaming trends aren’t just about catchy beats or technical features. They’re shaped by a mix of cultural trends, platform algorithms, listener habits, and how well a song connects with people across different places. This project helped me look past the charts and really understand what’s driving the music we see (and hear) on repeat.

--- title: " Understanding the Elements of Spotify Song Success " author: "Perry Lampertius" toc: true format: html: code-tools: TRUE embed-resources: TRUE execute: warning: FALSE message: FALSE --- ## Introduction Music plays a huge role in culture and daily life, and with platforms like Spotify and Apple Music, it's more accessible than ever. Millions of people stream music every day, but not every song gets the same amount of attention. Some tracks go viral and hit the top of the charts, while others stay under the radar. This made me wonder—what actually makes a song successful on a streaming platform like Spotify? Is it the tempo, how danceable it is, how energetic it sounds, or maybe even what day it was released? Do these kinds of musical and platform-related features really affect how popular a song becomes? As someone who enjoys music and regularly follows what’s trending on charts and playlists, I wanted to look into which features make a song more likely to become a hit. I think understanding these patterns could be helpful not only for music fans but also for artists, producers, and marketers who want to know what drives a song’s success in today’s streaming world. To explore this, I chose a dataset that tracks the performance and characteristics of songs across major streaming platforms, with a focus on Spotify. This dataset includes information like stream counts, playlist and chart appearances, and audio features such as tempo, energy, and mood. My goal is to analyze these variables and see which ones are most strongly connected to popularity. Can we find any trends or clues that explain why some songs take off while others don’t? That’s what I’ll be exploring in this project. ## Dataset Overview This dataset, (click here to download), offers a detailed snapshot of what drives song popularity across major music streaming platforms, including Spotify, Apple Music, Deezer, and Shazam. It brings together both quantitative performance metrics—such as total streams, playlist appearances, and chart positions—and a wide range of musical features like tempo (BPM), energy, danceability, valence (mood), and more technical aspects such as key, mode, and acousticness. Additionally, it includes contextual information like release dates and artist count, allowing for analysis over time or by artist collaboration scale. By combining musical characteristics with platform-specific performance data, the dataset provides a unique opportunity to explore not only which types of songs become popular, but also why they might succeed. For example, are high-energy songs more likely to be featured in popular playlists? Do slower tempos dominate certain seasons or genres? Does the number of artists on a track correlate with higher chart performance? This dataset can be used to examine trends in music consumption, identify the traits of viral or commercially successful tracks, and even inform recommendations for artists and producers looking to optimize their sound for streaming platforms. It’s also valuable for academic research, marketing insights, and anyone interested in the intersection of music, data science, and digital media trends. In an era where streaming data plays a significant role in shaping the music industry, datasets like this one offer a window into the mechanisms of musical success in the digital age. ## Key Variables in the Dataset ### Basic Track Info - `track_name`: Title of the song\ - `artist(s)_name`: Name(s) of the performer(s)\ - `artist_count`: Number of credited artists\ - `released_year`, `released_month`, `released_day`: Release date (year, month, day) ### Platform Performance - `streams`: Total number of streams (Spotify)\ - `in_spotify_playlists`, `in_apple_playlists`, `in_deezer_playlists`: Playlist count by platform\ - `in_spotify_charts`, `in_apple_charts`, `in_deezer_charts`, `in_shazam_charts`: Chart appearances per platform ### Musical Features - `bpm`: Tempo (beats per minute)\ - `key`: Musical key (e.g., C, D#)\ - `mode`: major key and minor key ### Audio Characteristics - `danceability_%`: Danceability score\ - `valence_%`: Positivity or happiness of the song - `energy_%`: Intensity level\ - `acousticness_%`: Degree of acoustic sound\ - `instrumentalness_%`: Likelihood of having no vocals\ - `liveness_%`: Likelihood the song was performed live\ - `speechiness_%`: Amount of spoken words ## Summary Table of Statistics ```{r} #| echo: FALSE #| library(dplyr) library(readr) library(knitr) library(tidyr) songs <- read_csv("https://myxavier-my.sharepoint.com/:x:/g/personal/lampertiusp_xavier_edu/ERAd5OYmzZVCoIy2h7SztcgBvzILHxqXGIW3G_5yAVtTUw?download=1") songs$streams <- as.numeric(gsub("[^0-9]", "", songs$streams)) summary_vars <- songs %>% select(streams, bpm, `danceability_%`, `valence_%`, `energy_%`, `acousticness_%`, `instrumentalness_%`, `liveness_%`, `speechiness_%`) summary_stats <- summary_vars %>% summarise(across(everything(), list( Mean = ~mean(., na.rm = TRUE), Min = ~min(., na.rm = TRUE), Max = ~max(., na.rm = TRUE), SD = ~sd(., na.rm = TRUE) ), .names = "{.col}_{.fn}")) summary_table <- summary_stats %>% t() %>% as.data.frame() %>% tibble::rownames_to_column("Metric") %>% rename(Value = V1) %>% tidyr::separate(Metric, into = c("Variable", "Statistic"), sep = "_(?=[^_]+$)") %>% mutate( Value = as.numeric(Value), Value = format(round(Value, 2), big.mark = ",", nsmall = 2, scientific = FALSE) ) kable(summary_table, caption = "Summary Table of Song Statistics", align = "lc") ``` The dataset features a wide mix of music, covering a broad spectrum of styles, tempos, and moods. Some tracks are fast-paced and high-energy, clearly designed for dancing or workout playlists, while others lean more acoustic and relaxed, likely intended for quieter, more introspective listening. This musical diversity makes the dataset rich for analysis, allowing comparisons across different genres and emotional tones. One of the most striking aspects is the dramatic difference in stream counts. While some songs have only a few thousand streams, others have reached into the billions. This shows that music streaming success is far from evenly distributed—a small group of massively popular songs captures the vast majority of listener attention. The high standard deviation in stream counts further confirms this imbalance, highlighting how just a few viral or globally successful tracks can heavily skew the data. These outliers are likely boosted by factors like playlist placements, artist fame, and algorithmic promotion. Looking at the musical features, there’s significant variation in metrics like BPM (tempo), energy percentage, and danceability. This opens the door for deeper exploration—such as clustering songs by genre, mood, or even by their likelihood of appearing on curated playlists. Despite the overall diversity, a common pattern emerges: most songs in the dataset are vocal-heavy and high in energy, while purely instrumental tracks are quite rare. This suggests that the dataset is tilted toward modern, mainstream pop music that’s crafted to be emotionally engaging and instantly catchy—music made to grab attention quickly in a fast-moving digital landscape. In short, the dataset offers a vivid snapshot of today’s popular music culture: bold, loud, energetic, and emotionally charged, with a strong commercial focus on what resonates with large audiences on streaming platforms. ## Visualizing the Trends Behind Spotify’s Most Popular Songs ### Distribution of Spotify Streams ```{r} #| echo: FALSE library(ggplot2) library(dplyr) library(readr) library(scales) songs <- read_csv("https://myxavier-my.sharepoint.com/:x:/g/personal/lampertiusp_xavier_edu/ERAd5OYmzZVCoIy2h7SztcgBvzILHxqXGIW3G_5yAVtTUw?download=1") songs$streams <- as.numeric(gsub("[^0-9]", "", songs$streams)) songs_clean <- songs %>% filter(!is.na(streams) & streams > 0) ggplot(songs_clean, aes(x = streams)) + geom_histogram(bins = 50, fill = "skyblue", color = "black") + scale_x_log10(labels = label_comma()) + scale_y_continuous(labels = label_comma()) + labs( title = " Distribution of Spotify Streams", x = "Number of Streams (log scale)", y = "Frequency" ) + theme_minimal() ``` This chart gives us a good look at how popular songs are on Spotify. Most songs don’t get a ton of streams, but a few really famous ones get massive numbers. That makes it hard to compare them on a regular chart, so using a log scale on the x-axis helps a lot. It basically evens things out so we can actually see the differences without the super popular songs completely overshadowing the rest. With the log scale, we can spot more patterns—like the fact that a lot of songs seem to land around the 100 million stream mark. That seems to be a sweet spot for songs that are doing really well, even if they’re not global hits. It shows there’s a big gap between your average song and the ones that completely blow up. Overall, this chart helps break down those differences and makes it easier to understand what success looks like on Spotify. ### Relationship Between Energy and Stream Count ```{r} #| echo: FALSE songs <- songs %>% mutate(streams = as.numeric(streams)) songs %>% mutate(streams = as.numeric(streams)) %>% ggplot(aes(y = `energy_%`, x = streams)) + geom_point(color = "#1f77b4", alpha = 0.7, size = 2) + scale_x_log10(labels = scales::comma) + labs( title = "Relationship Between Energy and Stream Count", x = "Number of Streams (log scale)", y = "Energy Percentage (%)" ) + theme_minimal(base_size = 15) ``` Looking at the data, there seems to be a bit of a connection between how energetic a song is and how many streams it gets, but it’s not super strong. A lot of the songs with around 75% energy do seem to have higher stream counts, so that could mean people are drawn to tracks that feel more upbeat or intense. Still, the trend doesn’t hold up across the board. There are plenty of high-energy songs that didn’t take off, and some chill, low-energy ones that ended up doing really well. So yeah, energy might help a song stand out or make it more playlist-friendly, but it’s definitely not the only thing that matters. Streaming success probably comes down to a bunch of different factors—like who the artist is, how the song gets promoted, whether it lands on a big playlist, and even timing. In the end, energy alone doesn’t explain why a song blows up, but it might give it a bit of an edge. ### Density of Streams by BPM ```{r} #| echo: FALSE songs %>% mutate(streams = as.numeric(streams)) %>% ggplot(aes(x = bpm, weight = streams)) + geom_density(fill = "#2ca02c", alpha = 0.6) + labs( title = "Density of Streams by BPM", x = "Beats Per Minute (BPM)", y = "Weighted Density (by Streams)" ) + theme_minimal(base_size = 15) ``` The peak in the graph indicates that songs within the 110–115 BPM range tend to receive higher stream counts overall, suggesting that this tempo is especially popular among listeners. This aligns with the fact that many mainstream pop and dance tracks fall within this range—fast enough to be energetic, but still easy to follow and accessible. However, it’s important to note that not every song in this tempo range is a hit. There are plenty of tracks with similar BPMs that haven’t gained much traction, and many successful songs fall outside of this range as well. Flatter areas of the graph show that songs at certain other tempos typically receive fewer streams on average, even if there are a large number of them. Because the density plot is weighted by stream count, just a few highly popular songs can create a noticeable peak, even if that tempo isn’t the most common overall. While this trend suggests that certain tempos may be more appealing to broad audiences, BPM alone doesn't determine a song’s success. A track's popularity is likely influenced by a mix of factors, including artist recognition, marketing, playlist placement, production quality, and cultural trends. Tempo may contribute to a song’s appeal, but it’s just one piece of a much larger picture. ### Stream Popularity by Mode (Major/Minor) ```{r} #| echo: FALSE library(scales) songs %>% mutate(mode = as.factor(mode)) %>% group_by(mode) %>% summarise(mean_streams = mean(streams, na.rm = TRUE)) %>% ggplot(aes(x = mode, y = mean_streams, fill = mode)) + geom_bar(stat = "identity", alpha = 0.7) + labs( title = "Stream Popularity by Mode (Major/Minor)", x = "Mode (Major/Minor)", y = "Average Stream Count" ) + theme_minimal(base_size = 15) + scale_fill_manual(values = c("blue", "orange")) + scale_y_continuous(labels = scales::comma) ``` This bar graph suggests that songs in a major key tend to receive more average streams than those in a minor key. In music theory, major keys are often associated with happier, brighter, and more uplifting emotional tones, while minor keys are typically perceived as more melancholic, introspective, or serious. From a commercial perspective, audiences often gravitate toward songs that evoke positive emotions, especially in mainstream pop, dance, and commercial radio formats. Major key songs may therefore enjoy broader appeal, making them more likely to be added to popular playlists, shared by listeners, or used in upbeat media and advertisements. These factors can lead to a positive feedback loop, where major key songs get more exposure and, as a result, accumulate more streams. Additionally, music production and songwriting trends often favor the major key for hit songs, partly because of its emotional impact and partly due to market-tested formulas. Artists and producers aiming for viral success might deliberately choose major keys for this reason. However, it's important to acknowledge that musical taste is diverse, and many successful songs are written in minor keys—especially in genres like hip-hop, alternative, R&B, or classical-influenced pop. Still, the aggregate data suggests a general preference or market trend favoring the emotional brightness of major-key songs in the streaming landscape. ### Correlation between Speechiness and Spotify Streams ```{r} #| echo: FALSE # Load required libraries library(dplyr) library(ggplot2) library(scales) # Convert columns to numeric if needed songs <- songs %>% mutate( `speechiness_%` = as.numeric(as.character(`speechiness_%`)), streams = as.numeric(as.character(streams)) ) %>% filter(!is.na(`speechiness_%`) & `speechiness_%` > 0) # Scatterplot of speechiness vs streams ggplot(songs, aes(x = `speechiness_%`, y = streams)) + geom_point(color = "orange", alpha = 0.6) + labs( title = "Correlation between Speechiness and Spotify Streams", x = "Speechiness (%)", y = "Spotify Streams" ) + scale_y_continuous(labels = comma) + theme_minimal(base_size = 15) ``` The analysis indicates that songs exhibiting moderate levels of speechiness—specifically in the range of 4–5%—tend to receive higher streaming counts compared to tracks with either minimal or excessive speech-like characteristics. This trend suggests a potential relationship between the incorporation of spoken-word elements and a song’s commercial appeal on platforms such as Spotify. In the context of audio feature analysis, speechiness refers to the extent to which a track contains spoken words. Lower values generally represent purely sung or instrumental compositions, while higher values are characteristic of tracks dominated by speech, such as spoken word recordings, skits, or a cappella rap. The observed peak in streaming performance at moderate speechiness levels implies that audiences may gravitate toward songs that balance traditional musical structure with subtle spoken or rhythmic vocal components. These may include: Conversational or narrative-style hooks, Rap verses integrated within melodic pop tracks, Spoken intros, ad-libs, or rhythmic vocal interjections, Such stylistic choices are particularly common in mainstream genres like pop, hip-hop, reggaeton, trap, and electronic dance music, where the fusion of melodic and rhythmic vocal delivery enhances listener engagement and emotional impact. This pattern may reflect consumer preferences for music that is energetic, dynamic, and relatable, as speech-like inflections often convey attitude, authenticity, or immediacy. Conversely, tracks with very high speechiness may lack the melodic elements that typically drive streaming popularity, limiting their resonance with broader audiences. ### Average Streams by Song Type (Solo vs Collaboration) ```{r} #| echo: FALSE library(ggplot2) library(dplyr) library(scales) songs %>% mutate( collaboration = ifelse(artist_count > 1, "Collaboration", "Solo"), collaboration = as.factor(collaboration) ) %>% group_by(collaboration) %>% summarise(mean_streams = mean(streams, na.rm = TRUE)) %>% ggplot(aes(x = collaboration, y = mean_streams, fill = collaboration)) + geom_bar(stat = "identity", alpha = 0.7) + labs( title = "Average Streams by Song Type (Solo vs Collaboration)", x = "Song Type", y = "Average Stream Count" ) + theme_minimal(base_size = 15) + scale_fill_manual(values = c("steelblue", "darkorange")) + scale_y_continuous(labels = comma) ``` The analysis shows that, on average, solo songs receive more streams than collaborative tracks, challenging the belief that collaborations automatically result in higher streaming numbers due to the combined fan bases. Several factors may explain this. Artist Popularity Bias is one key element. Influential solo artists, such as Taylor Swift or Drake, often have a dedicated following that drives their streaming success, regardless of collaborations. Their star power can overshadow any potential boost a collaboration might offer. Another factor is Diluted Audience Appeal. While collaborations aim to unite fan bases, they don’t always align in terms of audience tastes. Combining different styles or genres can lead to a product that lacks cohesion, potentially alienating listeners. Additionally, collaborations may not capture the same authenticity or creative focus that solo artists can achieve on their own. Marketing Power and Brand Identity also play a role. Solo artists have clear personal brands that resonate with their fans, and their promotions tend to be more focused. In contrast, collaborations often require joint marketing strategies, which can dilute the message and reduce impact. While collaborations can be strategic for expanding reach or exploring new ideas, this data suggests they don’t necessarily lead to greater streaming success. ### Average Streams by Month of Release ```{r} #| echo: FALSE library(ggplot2) library(dplyr) library(scales) songs %>% mutate( released_month = factor(released_month, levels = 1:12, labels = c("January", "February", "March", "April", "May", "June", "July", "August", "September", "October", "November", "December")) ) %>% group_by(released_month) %>% summarise(mean_streams = mean(streams, na.rm = TRUE)) %>% ggplot(aes(x = released_month, y = mean_streams, fill = released_month)) + geom_bar(stat = "identity", alpha = 0.8, width = 0.7) + labs( title = "Average Streams by Month of Release", x = "Month of Release", y = "Average Stream Count", caption = "Data sourced from the music streaming dataset" ) + theme_minimal(base_size = 16) + theme( axis.text.x = element_text(angle = 45, hjust = 1), axis.title = element_text(face = "bold"), plot.title = element_text(face = "bold", size = 18), plot.caption = element_text(size = 12, hjust = 1) ) + scale_fill_brewer(palette = "Set3") + scale_y_continuous(labels = comma) ``` The chart indicates that January exhibits the highest average number of streams among all months, suggesting that songs released at the beginning of the year tend to achieve stronger streaming performance on average. Several factors may contribute to this pattern. One potential explanation is seasonal listening behavior. Following the holiday season, many individuals resume regular routines—such as commuting, exercising, or working—which are often accompanied by increased music consumption. January may also present a less saturated release environment, as it typically sees fewer high-profile music releases compared to more competitive months like November and December. This relative lack of competition may provide songs released in January with greater visibility and a longer lifespan on playlists and charts. In addition, the start of a new year often prompts both streaming platforms and listeners to refresh playlists and explore new content, which may further enhance the exposure and streaming performance of songs released during this period. ## Comparison of Streaming Data with Billboard Top 100 Rankings After analyzing the dataset and identifying key variables that influence a song's popularity based on stream count, I would like to validate whether the current Top 100 Billboard rankings on Spotify align with the insights from the graphs created above. To do this, I will scrape data from the website https://kworb.net/spotify/, specifically collecting the daily Top 5 rankings from several countries over a two-week period. This timeframe will provide sufficient data to analyze which artists appear most frequently and how their popularity varies across different countries. I have chosen the following countries for analysis: United States, United Kingdom, Brazil, Germany, and France. These countries were selected because they represent large Spotify audiences with diverse musical tastes, providing a broad perspective on global streaming trends. ```{r} #| echo: FALSE library(rvest) library(ggplot2) library(dplyr) library(tidyr) urls <- c( "https://kworb.net/spotify/country/us_daily.html", "https://kworb.net/spotify/country/gb_daily.html", "https://kworb.net/spotify/country/de_daily.html", "https://kworb.net/spotify/country/br_daily.html", "https://kworb.net/spotify/country/fr_daily.html" ) scrape_top10_data <- function(url, country_name) { page <- read_html(url) table <- page %>% html_nodes("table") %>% html_table() data <- table[[1]] %>% select(Rank = 1, Track = 2, Artist = 3, Streams = 5) %>% mutate(Country = country_name) return(data) } all_data <- bind_rows(lapply(1:length(urls), function(i) scrape_top10_data(urls[i], gsub(".*/(.*)_daily.html", "\\1", urls[i])))) all_data_clean <- all_data %>% mutate(Streams = as.numeric(gsub(",", "", Streams)), Rank = as.numeric(Rank)) artist_count <- all_data_clean %>% group_by(Artist) %>% summarise(appearances = n()) %>% arrange(desc(appearances)) ``` ### Artist Distribution in Top 10 ```{r} #| echo: FALSE artist_distribution <- artist_count %>% filter(appearances > 4) print(artist_distribution) if(nrow(artist_distribution) > 0) { ggplot(artist_distribution, aes(x = "", y = appearances, fill = Artist)) + geom_bar(stat = "identity", width = 1) + coord_polar(theta = "y") + labs(title = "Artist Distribution in Top 10", fill = "Artist") + theme_minimal() } else { print("No artists with more than 5 appearances") } ``` The table shows 11 different songs and artists, and each one shows up exactly five times in the dataset. That means every song was popular in five different countries, but no one song or artist appeared more than the others. Even artists like Billie Eilish and Lady Gaga, who had more than one song on the list, still only showed up five times each. So, no one really stood out as being the most popular overall. This tells me that music tastes were pretty evenly spread out, and people in different countries were listening to a good mix of songs—not just a few big global hits. ### Total Streams for Top 5 Artists ```{r} #| echo: FALSE top5_artists <- all_data_clean %>% filter(Artist %in% artist_count$Artist[1:5]) %>% group_by(Artist) %>% summarise(total_streams = sum(Streams)) ggplot(top5_artists, aes(x = reorder(Artist, -total_streams), y = total_streams, fill = Artist)) + geom_bar(stat = "identity") + labs(title = "Total Streams for Top 5 Artists", x = "Artist", y = "Streams") + theme_minimal() + theme(axis.text.x = element_text(angle = 45, hjust = 1)) ``` The chart shows the total streams for the top five artists, and Jennie’s song “Like Jennie” is way ahead of the rest—almost three times more streams than the second artist. That kind of lead really makes her stand out in the dataset. The other artists still did well, but their numbers are a lot lower, which shows that Jennie’s song got way more attention overall. This big gap could mean Jennie’s track had a stronger global reach—maybe it got more playlist placements, more promotion, or just really clicked with people across different countries. Whatever the reason, “Like Jennie” clearly comes out on top and shows a level of popularity that the others didn’t quite match. ## Conclusion After going through all the data, one thing is super clear—there’s no exact recipe for what makes a song go viral on Spotify. Features like energy, BPM, or even the month it drops definitely matter, but they’re not a guarantee. A track can check all the boxes and still flop, while another song that breaks all the "rules" might end up everywhere. That tells me it’s not just about how a song sounds—it’s also about timing, marketing, who the artist is, and maybe even just being in the right place at the right time. One of the biggest things that stood out to me was how uneven streaming success is. A few songs rake in billions of streams, while most barely scratch the surface. Jennie’s “Like Jennie” was a perfect example of this—it totally dominated the top 5 and showed how one song can take off way more than the rest. But on the flip side, there were also artists like Billie Eilish and Lady Gaga whose songs were more evenly spread out across countries, showing that music taste can be really diverse depending on where you are. At the end of the day, the data shows that streaming trends aren’t just about catchy beats or technical features. They’re shaped by a mix of cultural trends, platform algorithms, listener habits, and how well a song connects with people across different places. This project helped me look past the charts and really understand what’s driving the music we see (and hear) on repeat.