Issue and Argument

With the dominant digitization of music and decreased barriers to entry, market competition in the music industry has surged; as of 2023, Spotify’s catalog exceeded 100 million songs (About Spotify, 2023). This new reality, combined with increasing market concentration (Dellatto, 2022; Aguiar & Waldfogel, 2016), compels musicians to seek alternative income sources, especially those who are not chart-toppers. Alongside other alternative sources of income such as the sale of merch (e.g., branded t-shirts), one of the avenues explored by musicians is to place their songs in advertisements and in other cultural products such as films, television, and—our focus here—video games.

For musicians, placing songs in video games is a gateway to a vast market. In 2022, the music industry revenue reached $26.2 billion (IFPI, 2023), dwarfed sevenfold by the video game industry’s $182 billion (Rousseau, 2023). Moreover, music has become integral to the gaming experience (Arevalo, 2023), with the unveiling of EA’s annual soccer game playlist (formerly named FIFA, now EAFC) as eagerly anticipated as the software itself (Face, 2021). Surprisingly, scientific attention to the effectiveness of song placements in video games is limited, leaving a important gap in the understanding of placement success. Yet, song placement is a unique form of product placement that deserve scientific consideration; contrary to other types of products, songs are placed in games both to advertise music and to contribute to the game experience. Hence, this study aims to develop and test a conceptual model (Figure 1) to assess song placement effectiveness in video games.

Conceptual model and hypotheses

Two factors may explain the main effect of song placement in video games on song performance. Firstly, video games are inherently enjoyable, immersing gamers in a pleasant, arousal-filled environment where favorable emotions can transfer from the game to the placed products (Ingendahl et al., 2022). Secondly, video games are designed to engage gamers for extended periods, possibly spanning weeks or months, with the same content. Consequently, a song integrated into a video game has the opportunity for repeated plays, enhancing its performance, as repetition is a crucial success factor in advertising (D’Hooge et al., 2017). In sum, we hypothesize that a song placed in a video game will benefit from the favorable context and frequent exposure typical of this medium, ultimately boosting song performance on other platforms like Spotify.

H1: Song placement in video games has a positive impact on song performance.

Secondly, we anticipate that song placement by familiar brands, i.e., popular and renowned artists, will prove more effective than placements by lesser-known brands. Indeed, studies on product placement in video games (Hwang et al., 2017; Jeong & Biocca, 2012) suggests that consumers more frequently recall products from familiar brands. Martí-Parreño et al. (2017) explain this phenomenon using Lang’s Limited Capacity Model of Mediated Message Processing (Lang, 2000); since video games are rich means of communication, players cannot effectively process all the available information and will focus primarily on the gameplay and then on other cues such as product placements. In this context, brand familiarity enables players to digest information conveyed during product placement without excessive cognitive effort, enhancing placement effectiveness. In the context of this study, we thus expect that when a gamer encounters a song by a well-known artist, information processing will be quick, facilitating the transfer of affect between the game and the song.

H2: The positive effect of song placement in video games on song performance is stronger for familiar brands (vs. unfamiliar brands).

Thirdly, we anticipate that the context of song placement will moderate its impact on song performance. Most prior studies on product placement in video games have concentrated on in-game advertising, assessing the effectiveness of a product placement during a virtual tennis match (Hwang et al., 2017) or within a shooting game session (Ingendahl et al., 2022). Interestingly, in the case of music placement, many songs are heard in contexts peripheral to the gameplay, like in a character selection menu, prompting us to question the differential effectiveness of in-game and out-of-game placement. In other entertainment media, placements with a high level of prominence, such as a product used by a movie’s main character, tend to be more effective than subtler ones, like a small banner ad in the background (Sharma & Bumb, 2020). In the context of song placements, we thus expect that their effectiveness is stronger when the song is deeply integrated into gameplay—during a crucial cutscene or a significant quest—and less effective when used in a peripheral context, such as a menu.

H3: The positive effect of song placement in video games on song performance is moderated by placement context. Hence, songs placed during the gameplay should benefit more from the placement than songs placed in a peripheral context.

Lastly, we expect that the impact of song placement is influenced by other songs simultaneously placed in the same game. In most games featuring licensed music, players encounter an extensive list of musical titles (for instance, FIFA 22’s soundtrack boasts 122 songs). Since we know that in sponsorship, a conceptually similar phenomenon to product placement, brands can gain from the positive associations linked to concurrent sponsors of the same sporting event (Boronczyk and Breuer, 2021), we hypothesize that a song placed in a game with a playlist featuring numerous popular artists will outperform a song placed alongside less popular ones. Indeed, by being placed alongside other popular artists, songs should benefit from the positive attitudes that gamers have about these well-known musicians.

H4: The positive effect of song placement in video games on song performance is moderated by concurrent songs’ popularity. Hence, songs placed in games which features other popular songs will outperform those placed alongside less popular ones.

Methodology

To test our conceptual model, we use a large secondary dataset, assembled from various data sources, which will be analyzed with linear growth models (see Singer, 1998). To create our sample, we identified an exhaustive list of video games published between 2017 and 2022 and that used licensed music (therefore song placements), as well as a list of all the songs placed in these games. Our preliminary sample contains 137 games and 6066 songs. For each of these songs, we collect data on their performance, i.e., the daily number of streams on Spotify (the main streaming platform in the music industry; Aguiar and Waldfogel, 2021) 3 months before and after the release of the video game. This approach allows for a precise assessment of the impact of the game’s release and, consequently, the effect of product placements on song performance. To measure brand familiarity, we use Spotify’s popularity index for each artist 3 months prior to the game’s release to mitigate potential endogeneity concerns. Additionally, to gauge the popularity of concurrent songs, we use the average popularity of all the artists featured in a game’s soundtrack. To assess placement context, we conduct manual coding, analyzing each game/song pair to determine whether the song was incorporated into gameplay or placed in a peripheral context. Furthermore, we will collect a range of control variables (game genre, game popularity, song genre, etc.), as outlined in Table 1.

Agenda and Stage of Completion of Research

The study sample has been identified, and our research team is currently in the midst of the data collection and coding phase. This includes coding the product placement contexts and gathering data from Spotify, Chartmetric, and IGDB APIs. We anticipate completing this phase by January 2024, after which we will proceed with data analysis during the winter of 2024.

Takeaway and Expected Contributions

Overall, this study will make significant contributions to the literature in three key ways:

1. It addresses an important phenomenon within the music industry that has been largely overlooked in scientific literature to date. Our dataset, which accurately measures song performance in the market, will also allow us to precisely quantify the effectiveness of song placements.

2. In the literature on product placement in entertainment products, results have shown significant context dependency, making it challenging to explain the observed variance (see Van Berlo’s 2021 meta-analysis on advertgames). By incorporating variables that measure the type of song placed and the placement context into our model, we aim to enhance our understanding of how to design the most effective product placements for artists.

3. By identifying the most effective contexts for product placements, our research will assist video game studios in optimizing the use of licensed music tracks, which can represent substantial costs. This will enable them to leverage these songs in contexts where gamers can fully appreciate their value.

Using a between-subject randomised experiment, this study examines perception bias related to AI creativity in music generation. Unlike previous studies, we remove the compositional aspect of the performance by exclusively focusing on AI piano cover songs, as examples of `musical style transfer'. Based on four well-known classic and contemporary pop songs, we find consistent evidence of bias against AI song performances when participants are informed the song is generated by AI versus when they (falsely) believe it is a human performance. We find this bias is greatest when related to the perceived ability of AI to perform a `creative interpretation' rather than a `good representation' of the original song. We also find some evidence that the bias may relate to familiarity with the original song, which may extend from a sentimental/emotional connection that participants hold. We find no evidence self-reported musical engagement or musical ability (based on two incentivised listening tests) have any relationship with bias. Finally, we confirm the existence of a perception bias by engaging a professional pianist to perform the sheet music generated by the AI reproduction.