To theme or not to theme: Can theme strength be the music industry's “killer app”?

Abstract

Music bundling has been the mainstay of the music industry for decades. Record companies and producers have selected bundles of songs and sold them as albums, their most important revenue source. Digitization and piracy of music have threatened this standard business model with consumers increasingly purchasing music a la carte. In this study, we analyze a strategy for designing successful albums through using new concepts of themed bundling. Thematic bundling can lower consumer search costs and dampen the incentive to pirate music, and can potentially be a win–win strategy for both consumers and music companies. Unlike prior work in economics on bundling which typically seeks to determine the optimal price, bundle size and composition, we focus on a restricted bundling problem, since the price of the bundled product (i.e. an album) is generally set over a narrow range as is the number of items (i.e. songs) in the bundle. Our key results and insights are derived using analytic modeling and extended through numerical analysis. In addition, our key findings are supported by our empirical analysis of music album chart performance.

Introduction

The album is dead! Long live the album! For centuries it was the town crier who first announced the death of the king and then proclaimed best wishes for the heir, the new king. In the music world, the album has long been the king, the major industry product. Album sales were the “measure of success” and artists were judged by the market performance of their albums, especially their most recent album.

But then came the rapid emergence of technology that provided the means for a new world of music. Consumers were freed from the shackles of albums, of music compilations forced upon them by the record companies. Individuals could now purchase sets of individual songs that they personally select. In July of 2007, the Wall Street Journal reported (July 5th, 2007) that while the 229.8 million albums sold in the first six months of 2007 represented a decrease of 15% over the same period in the previous year, individual digital track sales increased 49% to 417.3 million over the same period. Now surely the king was dead, and the individual song would be the new king! But is it the album that is dying or are different subsets of albums actually doing well and thriving? Is the album dying a slow death or could it be that, like many royal survivors of yore, there are successful forms of the album genre that have already or could quietly morph to assimilate and thrive in the new environment? Here we investigate whether theme strength (as defined in detail below) can be the basis for a successful album bundling strategy.

Today's music albums are bundles of digital goods. In the material that follows, we analyze the digital good bundling problem and address the following question: is it possible that major differences between successful album sets and unsuccessful album sets are linked to decision making in the bundling of digital goods? We develop formal models for the special digital good case of music. As we explain below, the bundling problem for digital goods is actually a very different problem from the bundling problem economists have studied extensively. In particular, we focus on a restricted bundling problem, since the price of the bundled product (i.e. an album) is generally set over a narrow range as is the number of items (i.e. songs) in the bundle. In the more general bundling literature, (see especially Geng et al. [10]), the models seek to determine the optimal price, bundle size and composition.

Nelson [14] classified goods into search goods and experience goods. Consumers can evaluate a search good prior to purchase by evaluating observable characteristics of the good together with the price of the good. An experience good, on the other hand, can only be evaluated through consumption or use. To accomplish this prior to purchase, consumers must be able to search and identify contending products and must have a sampling process available to them. Thus, consumers' search and sampling “experience” can both be important components of a consumer's individual estimation of his value for an experience good [17].

For many tangible experience goods, such as a brand of tuna, a consumer may need to sample and experience only once. If there is brand consistency for the product, the consumer can expect the taste from subsequent purchases of the brand to remain consistent. However, for experience goods such as music, movies, video games, software, or fiction books there can be high variability in quality (see Bhattacharjee et al. [2]) even across what might seem to be a “brand”. That is, a specific director's films often vary greatly as do the video games of a production team or the albums of a musical artist or the books of a fiction mystery writer. Liking one of the products in a product grouping (e.g., films by a specific director or books by a specific author) does not guarantee a favorable view of another product in that product grouping. That is, the elements of specific product groupings do not rise to the level of a “brand”. To make good choices, consumers must repeatedly search and sample for new offerings within a product grouping of interest to them. Further, the number of products of potential interest may be very large, presenting an imposing consumer search and sampling problem. Consider, for example, a consumer interested in folk music or mystery novels. How large are the sets that the consumer would need to search and sample in order to realize a high probability of a successful purchase?

Sellers are keenly aware of the benefit from offering consumers the ability to sample part of an experience good during the purchase decision process. Movie releases have long been preceded by trailers or “coming attraction” promos providing clips (samples) from the film. Software vendors often provide restricted edition access or limited 30-day free trials. Brick-and-mortar music retailers have provided various facilities for consumers to sample songs in-store before buying. Online music retailers provide a variety of brief or “partial song” sampling options. The Financial Times (website) now provides free access to the first two paragraphs of all news articles, requiring readers to subscribe for access to the rest of the article.

A key characteristic of music purchases is “purchase hazard” since music once purchased cannot be “returned” easily. For many physical experience goods, including clothes, returns are possible. In those cases, consumers have the safety net of getting their money back and thus avoiding loss, though this characteristic may lead to higher prices for such goods. Returns of music have been a bit trickier. Even when music is distributed on CD-ROM or other physical media, most retail stores allow exchanges but not returns of music CDs where the cellophane wrapping has been “breached”. In the case of digitally downloaded songs, the only way for a consumer to “return” a song would be to destroy the copy, which is not very reliable from the seller's viewpoint. Thus, for music, the avoidance of purchase errors is important since consumers cannot rectify such mistakes by returning the goods. Effective search and sampling can provide the means for consumers to make better purchase decisions.

The music recording industry has long played an active, or perhaps better termed a “pro-active” role in music sampling through payola, where payments of undisclosed sums are made to broadcast specific songs over radio or television. Coase [8] saw payola as a pricing system, where the service of broadcasting a record by a radio station increased the value of the record. That is, the act of broadcasting attracts a price (payola) which is a means of allocating a scarce resource (broadcast time). Payola can be viewed as a “forced” passive sampling mechanism that focuses sampling on a restricted music set. Though payola was banned by the FCC (Federal Communications Commission) in 1960, the practice continues today [9]. There is some evidence that P2P sharing networks are providing some of the benefits of payola by helping customers sample songs before buying [4]. Coase argued that, while the reason for banning payola was to minimize “deception” of consumers, the ban had unintended consequences including less competition among record companies (higher product prices), and additional monitoring expenses by regulators.

For music, the advent of the Internet included the emergence of technologies and channels that facilitated sampling at new levels. Peer-to-peer file sharing technologies (Napster, Gnutella, KaZaA, etc.) led to widespread downloaded music files for either retention (piracy) or sampling (as discussed in [2] and [4]). Online retailers such Amazon.com provide various digital downloadable samples of songs. Websites that have emerged for the purpose of promoting specific artists, groups of artists, or a generic class of “lesser-known artists” provide downloadable samples ranging from partial song clips to entire singles.

When digital products are available for downloading, both a substitution effect (retention of downloaded copy in substitution for purchase of the digital good) and a sampling effect (pre-purchase sampling tending to enhance sales of the digital good) can arise (see Gopal et al. [11]). For the case of music, Bhattacharjee et al. [3] find: (i) peer-to-peer sharing of music files results in pre-purchase sampling as well as piracy-related lost sales, and (ii) sharing activity provides a lead indicator to album performance on BillBoard charts. Chellappa and Shivendu [7] suggest that when the quality of a digital good is under-estimated, the sampling effect can actually dominate and help sales.

The emergence of digital goods has opened an expanse of possibilities for bundling. Digital goods can be easily bundled across sellers (e.g., record labels) and good categories. Consider the possible bundling of digital goods for use in an iPhone!

The enormity of the digital goods space and the possibility set for bundling pose daunting problems for those seeking to identify optimal bundling strategies or optimal pricing strategies (the focus of much of economists' research on bundling). But can we identify bundling issues that we can analyze and from which we can gain important insights? In the case of the restricted bundling problem for digital music, we argue that this is indeed the case.

Music has a long history of being sold in bundles or “albums”. Albums do tend to have some linking theme or characteristic – that is, the songs in the album are typically by the same artist, belong to the same genre, and were produced in a given year or time period. But some music albums (such as “compilation albums”¹) have stronger themes. It is such stronger themes that draw our focus here.

Because they possess a strong theme, we expect greater similarity among the songs in a compilation album than we would see on normal albums. Such similarity, in turn, can impact consumers' sampling behavior. If an individual likes one song in a compilation album, there is an increased likelihood (compared to the same occurrence in a regular album), that the individual will like the other songs in the compilation (See for example the recent study by Borreau et al. [6] which analyzes complementarity issues in the current digital music marketplace). Prior research has emphasized the importance of matching customized information goods to consumers' preferences (e.g. [15]). In our work, we suggest that theme strength can serve as a proxy for consumers' preferences.

While record companies/producers have driven such “compilation albums” (today's most prominent form of themed music bundles), other themed bundles are emerging in today's digital world. In fact, today's music bundles can be vendor-sponsored, independent service-agent sponsored, or user-driven. There are numerous examples of vendor-sponsored themed bundles in music, including albums of songs recorded at multi-artist events (e.g., “Woodstock: Three Days of Peace & Music”), Christmas albums, or albums featuring a one-time collaboration of two artists (Bruce Hornsby and Ricky Skaggs or Nat King Cole and Natalie Cole), and themed collections of various artists (“Various Artists – Anthony Rother Presents We Are Punks”, “Various Artists – 5 Years of Get Physical”). An interesting example of vendor-sponsored, individual sponsored, and service-agent sponsored themed bundles can be found in Apple iTunes' Playlists feature. The iTunes' Playlists are lists of songs that either Apple can compile (vendor-sponsored themed bundles) or users themselves can compile (user-driven themed bundles), which are then published on the Apple iTunes website. A wide variety of service-agent playlists are also available, containing recommendations from celebrities, entertainers, and critics. Any user can purchase some or all songs in a playlist. A user-created playlist (also called an iMix) can contain up to 100 songs, is published on the iTunes website for one year, and is searchable by any user. iMixes can be voted by other users using a five-star system.

A recent Gartner report [13] detailed the growing popularity of playlists and other consumer taste-sharing applications in online music purchases. In a different context, Ragno et al. [16] analyzed the similarity of a list of songs played together (in one session) by a radio station. The researchers found various groupings that captured similarity among these songs – e.g. membership in the same genre, similar popularity, or suitability to be played together. The researchers used multiple such lists to build an algorithm that inferred similarity among songs and automatically generated similarity-based playlists. Thus there are two differing approaches, one driven by the record company and another by external entities, which currently determine the ex ante theme of the bundle. Here we have not modeled this ex ante decision making process.

A significant benefit of aggregating songs by theme is the likelihood of lowered search costs for consumers. Given the large and ever-growing collection of music, the recording industry is grappling with the challenge of helping consumers find the songs they like (and that they will buy!) in a quick and effective manner. Themed bundles can be one way to do this. Consumers who sample compilations possessing “high thematic strength” might quickly find music that fits their taste, without actually having to sample every individual song from a vast and ever-growing pool. Themed bundles of songs thus can serve the purpose of most recommender systems – to help “tell consumers about music that they might like”.

We approach the development of our formal model in a way quite different from what most economists have done in the long history of research on bundling. The prior stream of work in economics has tended to focus on identifying optimal pricing strategies for bundled goods (see, especially, discussions in Geng et al. [10]). Historically, music albums have fallen within a fairly tight price range and thus we take price as given (i.e., determined for an album). We focus on the issues of theme strength and sampling.

In Section 2, we offer a formal model of a consumer's music sampling and purchase decision process. The results of a simple analytical modeling exercise suggest that as theme strength increases and sampling costs decrease, consumers are more likely to purchase such an album compared to one with lower theme strength, even when the ex ante valuation is identical for the two albums. To the extent that sampling increases the likelihood of purchase, compilation albums possess both a competitive edge on regular albums and may result in less piracy. We substantiate the findings of this simple analytical model with numerical analysis of a more complex model in Section 3. In Section 4, we provide initial validation of the model findings using empirical evidence. We conclude in Section 5 with a discussion of key insights and managerial implications.