Article

Angélica Dolores Pereira de Melo; Paulo Fernando do Nascimento Afonso

DOI: 10.22167/2675-6528-2026M10

Article prepared by the ResumeAI tool, an artificial intelligence solution developed by the Pecege Institute focused on synthesis and writing.

Summary

Hit Song Science (HSS) explores techniques for predicting and classifying musical hits. The present study aimed to investigate the patterns of number 1 hits on the Billboard Hot 100, from 1958 to 2024, and propose variables for supervised machine modeling. To this end, an agglomerative hierarchical clustering machine learning technique was applied. The data collection covered 68 annual Billboard hits, considering intrinsic metrics such as beats per minute (BPM), energy, danceability, valence, and title repetition in the lyrics, as well as extrinsic characteristics such as the performer’s gender and collaboration format. The analysis revealed a significant increase in the longevity of hits from the 2000s onwards, the predominance of male voices, and the growing importance of partnerships between artists. Seven clusters of hits with distinct musical and lyrical characteristics were identified, and nine rules for current hits were proposed, based on the patterns observed in the sample. The results contribute to the HSS field, providing guidelines and variables for the development of predictive machine learning models.

Keywords: Data analysis; Clustering; Machine learning; Popular music; Musical patterns.

1. Introduction

The music industry, since its more formalized origins, has been driven by the pursuit and production of commercial successes. In the 19th century, Tin Pan Alley in Manhattan emerged as a hub where music was conceived and distributed with an industrial logic, characterized as song “factories” (Bishop, 1958). In this context, songs were treated as commodities, subject to manufacturing, promotion, and distribution processes, following capitalist principles of division, specialization, and standardization (Suisman, 2009).

The metrication of hits was fundamental to the dynamics of musical production. The sale of sheet music, although initially not very profitable, established the industry’s organizational structure, where substantial profit came from successes, or “Hits” (Suisman, 2009). This logic not only shaped the business but also influenced the very form of popular music, contributing to the emergence and standardization of genres such as jazz and blues (Suisman, 2009).

Over time, the music industry has evolved through various technological transformations, from the phonographic era of albums to the current streaming format, which refers to the sale of low-cost singles. Faced with this complexity and the constant search for success patterns, the field of Hit Song Science (HSS) emerged. The term HSS was introduced in 2003 by the artificial intelligence company Polyphonic HMI, which developed a hit prediction software used by renowned artists (Seufitelli et al., 2023).

HSS is a multidisciplinary field that integrates computer science with areas such as music theory, music sociology, and cultural markets (Seufitelli et al., 2023). Its main objective is to collect and analyze musical data from various modalities and sources, employing Information Retrieval, Machine Learning, and Data Mining techniques to predict a song’s potential success before its release. Understanding the structure of a hit is valuable for various market agents, including executives, digital service providers, and artists, aiding in investment maximization and strategic decision-making (Seufitelli et al., 2023).

Despite the relevance of HSS, previous research has predominantly focused on supervised techniques and the “hit or non-hit” binarity, often using data already organized by rankings, such as Top Charts (Seufitelli et al., 2023). This approach limits the exploration of intrinsic and extrinsic patterns in more diverse datasets and over extensive periods, leaving a gap in the understanding of the evolution of hit characteristics.

In this context, the present research adopts clustering, an unsupervised cluster analysis technique, to diagnose similarities and dissimilarities in a comprehensive dataset of hits. This choice is justified by the diversity of the data, which covers a period from 1958 to 2024, encompassing multiple decades and distinct cultural and musical identities. Clustering allows for the identification of underlying patterns that Tin Pan Alley publishers already intuited, but which can now be explored with scientific rigor (Fávero et al., 2024).

The study, therefore, applied machine learning techniques to identify patterns and cluster the 68 number-one hits on the Billboard Hot 100 from 1958 to 2024. The metrics explored by an agglomerative hierarchical method were the results of the analysis of these hits from “Top Chart Here’s Every Billboard Hot 100 Year-End No. 1 Song”, also considering the literature on the history of the music industry and Tin Pan Alley hits. Thus, the objective of this study was to investigate the patterns of hits from the phonogram era and propose variables for supervised machine modeling.

2. Material and Methods

The study adopted a quantitative and exploratory approach, employing machine learning techniques to investigate patterns in musical hits. The objective was to diagnose similarities and dissimilarities in a comprehensive set of hits, as well as to propose variables for future supervised machine modeling, as outlined in the introduction. The research focused on analyzing the intrinsic and extrinsic characteristics of songs that reached the top of the charts, following the taxonomy proposed by Seufitelle et al. (2023).

The unit of analysis consisted of 68 hits that reached the number one position on the Billboard Hot 100 annually, covering the period from 1958 to 2024. The data were collected from “Top Chart Here’s Every Billboard Hot 100 Year-End No. 1 Song”, available on the official Billboard website. This selection allowed for a longitudinal analysis of the hits’ characteristics over multiple decades.

For the collection of intrinsic characteristics of the hits, which refer to attributes of the audio itself, the tools Sort Your Music and chosic.com were used, both integrated with Spotify. These platforms allowed the extraction of metrics such as beats per minute (BPM), energy, danceability, valence, duration, key, and tempo. Such data were fundamental for understanding the musical properties of each hit.

The characteristics related to the lyrics and lyrical structure were obtained through the website musixmatch.com, synchronized with Spotify. Information was collected on the repetition of the title in the lyrics and the presence of elements such as intro, pre-chorus, chorus, verse, bridge, hook, and outro. Additionally, extrinsic characteristics, such as the performer’s gender (male, female, mixed) and collaboration format (solo, group, duet), were recorded.

For the year 1997, two Elton John hits, “Candle in the wind” and “Something about the way you look tonight”, were considered separately due to the absence of a single ISRC on Spotify that merged the two songs. Each was treated as a distinct phonogram. Furthermore, two instrumental hits, “Theme from a Summer place” (1960) and “Stranger on the shore” (1962), were excluded from the descriptive analyses of song form and title repetition in the lyrics, as they lacked these elements.

Regarding the interpreter’s gender, for mixed groups like “Ace of Base” and “Wings”, the gender of the one singing the verses was considered. In cases of featured artists and collaborations, where both artists shared parts of the verses, the interpreter’s gender was classified as “mixed”. This approach aimed to capture the diversity of vocal presentation formats present in the sample.

The data analysis employed the agglomerative hierarchical clustering technique, implemented in Python. Cluster analysis, which originated in Anthropology and Psychology (Reis, 2001; Fávero et al., 2009), was chosen to identify patterns and groupings of hits based on their similarities and dissimilarities. This exploratory technique is useful for verifying the existence of similar behaviors among observations (Fávero et al., 2024), being applied to diagnose the similarities and dissimilarities of the hits in the sample.

The metrics used for clustering were BPM, energy, valence, danceability, and title repetition. Before applying the linkage method and distance, title repetition was converted to seconds, and all metrics were standardized by Z-score. The selected dissimilarity measure was Manhattan distance, and the linkage method applied was Complete, for presenting greater diversity among the clusters.

The determination of the clusters was performed based on a cutoff measure of eight in the Manhattan distance, resulting in the formation of seven distinct clusters. This approach allowed for the grouping of the 68 hits into categories that reflect shared musical and lyrical characteristics, contributing to the identification of the patterns sought by the study.

3. Results and Discussion

The analysis of Billboard Hot 100 number-one hits, spanning the period from 1958 to 2024, revealed significant patterns in the longevity and intrinsic and extrinsic characteristics of the songs. Billboard’s primary criterion for classifying a hit as an annual number one is its permanence among the 100 best-selling and most-listened-to songs over 12 months. This longevity is a crucial indicator of a song’s ability to drive the music market, reflecting its impact beyond simple listening or sales, influencing multiple other businesses in the industry. The findings provide an empirical basis for understanding the dynamics of musical success.

It was observed that, of the 68 hits analyzed, the average number of weeks on the Top 100 was 26 weeks. The hit with the shortest longevity was “Bridge Over Trouble Water” (1970), which remained for 17 weeks, while “Heat Wave” (2022) demonstrated the greatest durability, with 91 weeks on the charts. Curiously, four hits were classified as annual number ones not for having reached the first position in their respective year, but for their notable permanence in the Top 100. Examples include “Surfin’ U.S.A.” (1963), which reached third position and remained for 25 weeks, and “Levitating” (2021), which reached second position and stayed for 77 weeks, evidencing the importance of consistency over time.

A remarkable behavior emerged when comparing the longevity of hits across different periods. Between 1958 and 1999, the average stay of a hit in the Top 100 was 25 weeks. However, in the period from 2000 to 2024, this average almost doubled, reaching 49 weeks. This 98% increase in the ability of a hit to remain relevant in public taste in the 21st century suggests a significant transformation in the dynamics of the music industry. Such a change can be attributed to technological evolution and more precise metricization of music consumption, which shifted from predominantly analog data to real-time quantification via streaming platforms, as discussed by Suisman (2009) regarding the logic of Tin Pan Alley.

The comparison of the longevity of iconic artists also illustrates this shift. While The Beatles, with two number-one hits in the 1960s (“I Want to Hold Your Hand” and “Hey Jude”), accumulated 34 weeks on the Top 100, singer Adele, with only one hit (“Rolling in the Deep”) in 2011, remained for 65 weeks. George Michael, with two hits in 1985 and 1988, totaled 43 weeks, and Elton John, with two hits in 1986 and 1997, reached 65 weeks. Paul McCartney, with three hits in different years, accumulated 53 weeks. This data, although it does not imply that current artists are listened to more than those of the 20th century, reflects the greater capacity for quantification and the control that contemporary artists have over the distribution and consumption of their music, driven by the digital age and streaming.

The 1990s represented a crucial transition period, with the average stay of hits rising to 37.3 weeks, in contrast to the 21.15 weeks of previous decades (1958-1989). This growth paved the way for the digital revolution of the 2000s. The influence of MTV and the popularization of Compact Discs (CDs) contributed to the increase in Billboard’s reports and revenues, diversifying forms of musical consumption. The longer lifespan of hits in the 21st century, with almost double the average of the 20th century, indicates that the industry reinvented itself, defying the pessimism of the late 1990s and demonstrating sustainability in the digital age, with greater control and democratization of distribution channels for artists. 21st century, the era of collaborators?

The analysis of the extrinsic characteristics of hits revealed an increasing trend of collaborations between artists in the 21st century. In the Tin Pan Alley era, the characteristics of hits were predominantly acoustic and genotypic, focused on musical and lyrical structure. With the recording industry, the performer gained prominence, and the artist’s popularity began to influence the value of the recording. Between 1958 and 1999, only three annual number-one hits were collaborations, representing 7.14% of the total. However, between 2000 and 2024, the number of partnerships increased to six, corresponding to 24% of the hits, a relative growth from 7% to 24% in a significantly shorter period of time.

This increase in collaborations can be interpreted as a strategy for hit-making in the contemporary era, where the diversity of artists and market competitiveness are greater. Partnerships allow for the expansion of fan bases and the exploration of different musical styles, which can be a differentiator for a song’s success. Although the total sample of 68 hits shows a balance between hits from groups/duos/partnerships (51.47%) and solo artists (48.53%), the analysis by decade reveals that, in the 21st century, solo artists have reached the number one position more frequently, despite the increase in collaborations. In the 1960s, seven hits were by groups, while from 2000 to 2024, only four hits were by groups, indicating a complexity in the success dynamic. The male predominance

The research identified a predominance of male performers among the 68 hits analyzed, representing 64.71% of the total. To contextualize this predominance, it is essential to consider the music industry’s landscape and cultural issues over the decades. The average number of annual number-one hits performed by women was 2.83 per decade. In the 1960s, only one woman was among the top ten annual hits. The 1970s and 2010s recorded three women, with two being solo artists and one a mixed hit in the 2010s. In the 1980s and 1990s, five women (four solo and one mixed hit) reached the top, and from 2000 to 2010, there were four women (three solo and one mixed hit).

The male predominance can be partially explained by Irving Berlin’s (1913) historical rule, which suggested that a hit should be within the average voice range so that the public could sing along. Although this rule originally applied to sheet music, it seems to have influenced the appreciation for male voices in the industry. However, the era of phonograms allowed singers’ technical expertise to become a differentiator. Female artists such as Barbra Streisand, Whitney Houston, Mariah Carey, Adele, Beyoncé, and Faith Hill, who feature among the number-one hits, often deviate from the standard of “average voice with not-too-high notes,” demonstrating that vocal talent can overcome this historical rule, although they still represent a minority in the sample. The pitches

The analysis of tonalities, an intrinsic characteristic of the hits, revealed that the key of C and its variations (C major, C minor, and C#) was the most frequent in the sample, present in 16 hits. The key of F and its variations (F major, F minor, and F#) was the second most common, with 12 hits. Of these, only two hits in C were sung by women (Roberta Flack in 1972 and Mariah Carey in 2005). The region between E and F (including the sharp variation of E, which is F) was the most sung in female hits, with four hits in E and four in F. In contrast, male hits concentrated on lower keys, in the regions of C and D and their variations.

Specifically, 11 hits in C were sung by men, two by women, two mixed hits, and one instrumental. In D and its variations, seven hits were sung by men, two by women, and one instrumental. The key C# (or Db) was present in seven hits, all sung exclusively by men. In total, 23 hits featured sharp keys, with 20 sung by men and only three by women. The keys C and F were the most frequent, at 23% and 17.65% respectively, both predominantly male. Regarding modality, 75% of the hits were in Major key, and only 25% were in Minor key, indicating a preference for more “positive” or “brighter” sounds in successful hits. Genotypic Characteristics Diagnosis of Hits

To deepen the understanding of success metrics, the research adopted the taxonomy by Seufitelle et al. (2023), which distinguishes intrinsic (genotypic) and extrinsic (phenotypic) characteristics of hits. Intrinsic characteristics refer to the audio itself, such as BPM, energy, danceability, valence, duration, key, and are considered “acoustic characteristics”. Extrinsic characteristics, on the other hand, are non-structural information, such as number of streamings, artist popularity, sales, and collaborations. Data collection for the intrinsic metrics of the 68 Billboard hits was performed using tools like chosic.com, which is based on Spotify data, allowing for a detailed analysis of sound art.

The average beats per minute (BPM) of the 68 hits was 123. The hit with the highest BPM was “Last Night” (Morgan Wallen, 2023), with 204 BPM, while the lowest was “I Will Always Love You” (Whitney Houston, 1993), with 68 BPM. The Energy metric, which measures the perceived intensity of the music, had an overall average of 58 (on a scale of 0 to 100). “Believe” (Cher, 1999) was the most energetic hit, with 92, and “The First Time Ever I Saw Your Face” (Roberta Flack, 1972) was the least energetic, with only three. A trend was observed for post-2000s hits to maintain the Energy level above 50, suggesting a preference for more intense music in the digital age.

Danceability, which indicates the ease of dancing to the music (scale from 0 to 100), had a total average of 63. “Low” (Flo Rida feat. T-Pain, 2008) was the hit with the highest danceability, reaching 92, while “Bridge Over Troubled Water” (Simon & Garfunkel, 1970) presented the lowest, with 15. The 1980s and 2010s stood out as the most danceable decades in the sample of hits. Valence, which measures the music’s ability to evoke positive or negative emotions (scale from 0 to 100), had an average of 60.82. Three hits tied for the highest valence of 97: “Surfin’ U.S.A.” (The Beach Boys, 1963), “Sugar, Sugar” (The Archies, 1969), and “Joy to the World” (Three Dog Night, 1971). The hit with the lowest valence, and the most melancholic, was “I Will Always Love You” (Whitney Houston, 1993), with 11.

The average duration of phonograms was 3 minutes and 50 seconds. The longest hit was “Hey Jude” (Beatles, 1968), with 7 minutes and 6 seconds, and the shortest was “Wooly Bully” (Sam the Sham and the Pharaohs, 1965), with 2 minutes and 22 seconds. From the 2010s onwards, a stabilization of hit durations between three and four minutes was observed. This standardization may reflect the demands of streaming platforms and listener attention in the digital age, where conciseness and retention are important factors for success. The intrinsic characteristics, therefore, offer a detailed overview of the musical attributes that contribute to the popularity of hits over time. The song form

The analysis of letter shapes and the temporal structure of hits revealed consistent patterns. It was found that 93% of the hits followed the 4/4 time signature pattern, while only seven hits were in 3/4 time. This predominance of the 4/4 time signature suggests a preference for more stable and conventional rhythms in popular music. All analyzed hits featured verses in their structure, and only two hits, “The First Time Ever I Saw Your Face” (Roberta Flack, 1972) and “The Way We Were” (Barbra Streisand, 1974), did not have a chorus. For the remaining 66 hits with lyrics, the average number of verses contained was 2.6, and the average chorus repetition was 2.8 times.

Other structural elements were examined. The “Introduction” was present in 24.24% of the hits, while the “Outro” (or Finalization), a distinct final section that is not a verse or chorus, was observed in 72.73% of the hits. This outro, which often features a different melody or lyrics than the rest of the song, suggests the importance of closing the lyrical narrative in a unique way. The “Pre-chorus”, a repetitive introduction to the chorus but without the same impact, was found in 25.76% of the hits. The “Bridge”, a section of variation between the verse and chorus that reinforces the central idea of the song, was present in 46.97% of the hits. Finally, the “Hook”, a short phrase or melodic/conceptual riff, was identified in 9.09% of the hits.

The repetition of the title, in whole or in part, throughout the lyrics was a striking feature in the 66 hits with lyrics, with an average of 17.95 times. This repetition, which can be explicit or implicit, as in “Boom Boom Pow” (The Black Eyed Peas, 2009), where the word “Boom” is reinforced, or “Every Breath You Take” (The Police), with the repetition of “Every”, demonstrates the importance of fixing the central idea of the song. A notable increase in title repetition was observed from the 2000s onwards, which may be related to the need for greater memorization and recognition in a more fragmented and faster music consumption environment, where immediate content identification is crucial for listener engagement. The formation of clusters of the 68 Hits

The application of agglomerative hierarchical clustering, using the Manhattan dissimilarity measure and the Complete linkage method, allowed the identification of seven distinct clusters among the 68 hits. The metrics of BPM, Energy, Valence, Danceability, and Title Repetition (converted to seconds and standardized by Z-score) were used to diagnose similarities and patterns. The “Complete” method was chosen for presenting greater diversity among the clusters, with a cut-off measure of eight in Manhattan distance. This exploratory technique was fundamental for grouping hits with similar characteristics, revealing the “formulas” that Tin Pan Alley editors already intuited.

Cluster One, named “The force of rap”, included hits like “Low” (Flo Rida feat. T-Pain, 2008), “Yeah!” (Usher feat. Lil Jon & Ludacris, 2004), and “Boom Boom Pow” (The Black Eyed Peas, 2009). These hits are predominantly from the rap genre and were released in the 2000s, sharing musical and lyrical characteristics that make them similar. Cluster Two, “The danceable ballads of different eras”, brought together songs like “Every Breath You Take” (The Police, 1983), “Macarena” (Los Del Rio, 1996), and “Uptown Funk” (Mark Ronson feat. Bruno Mars, 2015). Despite the diversity of eras, the similarity between these hits lies in their danceability, making them icons in dance clubs and parties over time.

Cluster Three, “Positivity in the post-war world. The American dream”, was composed of seven specific hits from the 1960s, including “Theme from a Summer Place” (Percy Faith and His Orchestra, 1960), “California Dreamin’” (The Mamas & The Papas, 1966), and “Surfin’ U.S.A.” (The Beach Boys, 1963). The lyrics of these hits express positive sentiments and jingoism, reflecting the “American dream” of the period. Cluster Four, “Positivity of the current world”, grouped hits predominantly from the 1990s and 2000s, such as “Hanging by a Moment” (Lifehouse, 2001), “Believe” (Cher, 1999), and “Blinding Lights” (The Weeknd, 2020). The lyrics of these hits emphasize feelings of self-confidence and self-esteem, as well as positive relationships.

Cluster Five, “Visual Hits”, was the most heterogeneous, including “Old Town Road” (Lil Nas X feat. Billy Ray Cyrus, 2019), “We Belong Together” (Mariah Carey, 2005), and “Rolling in the Deep” (Adele, 2011). This cluster is characterized by good-humored hits, songs with narrative music videos and/or lyrics that tell stories, suggesting that visual appeal and narrative are important elements for the success of these songs. Cluster Six, “Forever Romantics”, was one of the most homogeneous, with hits like “Stranger on the Shore” (Mr. Acker Bilk, 1962), “I Will Always Love You” (Whitney Houston, 1993), and “Candle in the Wind” (Elton John, 1997). The romantic character of all the songs is the main similarity.

Finally, Cluster Seven, named “Ballads,” was formed by hits such as “Hey Jude” (The Beatles, 1968), “Silly Love Songs” (The Wings, 1976), and “Careless Whisper” (Wham!, 1985). This cluster encompassed ballads from the 1960s, 1970s, 1980s, and 1990s, indicating that the ballad genre, with its specific melodic and lyrical characteristics, maintained its relevance as a number-one hit across several decades. The formation of these clusters demonstrates that, despite temporal and stylistic diversity, there are underlying patterns that can be identified and used to predict a song’s potential for success, contributing to the field of Hit Song Science. The nine rules for current hits, according to an analysis of Billboard’s Hot 100 and the trend of 21st-century hits for machine learning

Based on the analysis of Billboard Hot 100 hit patterns and observed trends in the 21st century, nine rules were proposed for training supervised machine learning models. Firstly, hits should have, on average, two verses, repeat choruses three times, and include an outro. Secondly, the predominant key should be major, and the ideal time signature is 4/4. Thirdly, hits should be sung in keys from C to E, within a medium vocal range, preferably by solo artists, facilitating audience identification and singing.

Fourth, the song title should be repeated in the lyrics, not just in the chorus, but throughout the structure, as a reinforcement of the central idea, with an average of 17 repetitions. Fifth, for hits sung by women, outside the average vocal range, the key of G is the most indicated, suggesting an adaptation to outstanding female vocal characteristics. Sixth, the average beats per minute (BPM) should be 123, indicating a rhythm that aligns with most hits. Seventh, it is crucial to invest in partnerships and collaborations between artists, given the growing trend of collaborative hits in the 21st century, which expand reach and fan base.

Eighth, hits should have an average duration of four minutes, reflecting the stabilization observed in recent decades and optimization for consumption on digital platforms. Finally, ninth, Valence, Energy, and Danceability metrics should vary around 60. These rules, derived from the empirical patterns of Billboard number one hits, provide concrete guidelines for the development of predictive models in Hit Song Science. They synthesize the characteristics that, over time, have proven consistent in musical success, offering a framework for future research and practical applications in the industry.

In summary, the investigation of Billboard Hot 100 number-one hits between 1958 and 2024 revealed a notable evolution in song longevity, with a significant increase in the digital era, and highlighted the growing importance of collaborations. The predominance of male voices, the most frequent keys in C and F, and the average BPM, energy, danceability, and valence around 60, along with structural patterns such as title repetition and the presence of outros, configure a set of distinctive characteristics. Clustering identified groups of hits with musical and lyrical similarities, allowing for the formulation of nine rules that serve as fundamental variables for the development of supervised machine modeling, contributing to the understanding and prediction of musical success in the field of Hit Song Science.

4. Conclusion

This study investigated the patterns of Billboard Hot 100 number-one hits from 1958 to 2024, aiming to propose variables for supervised machine modeling in the field of Hit Song Science. A significant increase in hit longevity was observed starting from the 2000s, with the average chart duration nearly doubling compared to the 20th century. A predominance of male performers and a growing trend of artist collaborations were noted, especially in the 21st century. Analyses revealed that C and F keys were the most frequent, with 75% of hits in major keys. In terms of intrinsic characteristics, the average beats per minute was 123, while energy, danceability, and valence orbited around 60. The average duration of phonograms stabilized between three and four minutes from the 2010s onwards. Regarding lyrical structure, the prevalence of 4/4 time signature, the presence of verses and choruses in most songs, and a notable repetition of the title throughout the lyrics, intensified in the 2000s, were identified. The application of agglomerative hierarchical clustering allowed for the identification of seven distinct hit groups, each with specific musical and lyrical characteristics, such as “The power of rap” and “Forever romantic,” evidencing underlying patterns that shaped musical success.

The main contribution of this work lies in the formulation of nine rules for current hits, derived from observed empirical patterns, which serve as essential guidelines and variables for the development of predictive machine learning models in Hit Song Science. These rules cover aspects such as lyrical structure, tonality, vocal range, title repetition, investment in collaborations, and audio metrics, offering a practical framework for the music industry and academic research. However, it is important to recognize that the direct comparison of music popularity and consumption between artists from different eras, such as those from the 20th and 21st centuries, is limited by the evolution of metrification methodologies, which have shifted from predominantly analog data to real-time quantification via streaming. For future studies, it is suggested to validate these rules on broader datasets and deepen the investigation into how behavioral changes in music consumption and new technological formats influence the longevity and success of hits, exploring the causality of these associations.

Bibliographic References

Berlin, Irving. 1916. Love-Interest´s as a Commodity. Green Book Magazine, Vol 13. Apr., pg. 695.

Bishop, W. 1958. How Popular Song Factories Manufacture a Hit. NYT, Sept. 18; “Reminiscences of Walter Bishop”, Sept. 1958, p. 23, CU-OHROC; Goldberg, Tin Pan Alley, 100.

Fávero, Luiz Paulo; Belfiore, Patrícia. (2024). Manual de análise da dados: estatísticas e machine learning com Excel, SPSS, R e Python. 2 ed. Rio de Janeiro: LTC.

Seufitelli, D.S.; Gabriel P. Oliveira; Mariana O. Silva; Clarisse Scofield and Mirella M. Moro. 2023. Hit song science: a comprehensive survey andresearch directions. In Journal of New Music Research. VOL. 52, NO. 1, 41–72. https://doi.org/10.1080/09298215.2023.2282999

Suisman, D. 2009. Selling sounds: the commercial revolution in American music. Havard University Press, Cambridge, Mass.

Article originating from the Final Project of the Specialization in Data Science and Analytics of the MBA USP/Esalq

_{Learn more about the course; click here:}