Pricing Strategies in the Market for Online Music

Pricing Strategies in the Market for

Online Music

A conjoint-based approach to bundling and versioning

Pricing Strategies in the Market for

Online Music

A conjoint-based approach to bundling and versioning

Student: Roderick Vos Student Number: 1468073

Supervisor: Dr. J.E.M. van Nierop Second Supervisor: Dr. Sonja Gensler

Date: April 2011

Faculty: Economics and Business Course: MSc. Business Administration

Profiles: Marketing Management & Marketing Research

Van Hamelstraat 33A 9714 HJ Groningen 0615053143

Summary

In this thesis pricing strategies for the market for online music are analyzed. Currently most online music transactions concern only the trade-off between a certain amount of money and, in return, one or more music files (be it in MP3 or some other format). Players in the market for online music have shown concern that this approach may not be the right one when it comes to addressing consumers with differences in preferences. Differences in consumer preferences seems to be a concept that is so straightforward and clear that it should be one of the basic principles when performing any type of business in any market. However, in the market for online music the current product and pricing strategies have been rather stale and traditional. As a result of this there have been signals in the market that things should be done differently. One of the ways the strategies could be altered is through versioning and bundling. These are different approaches to the current pricing strategies in the market. Versioning entails offering different product versions (e.g. by inclusion of extra features into the product) to address different types of customers, each with different concepts of value. Bundling concerns the offering of different packages of complementary products. Consumers have a choice of paying more to receive more. Previous academic studies have shown that offering a product specific to one’s needs should increase one’s perceived value. This in return should have a positive effect on people’s intention to purchase and their willingness to pay.

Table of Contents

1. Introduction and Research Questions ... 5

1.1 Introduction... 5

1.2 Different Business Models in the Online Channel ... 7

1.3 New Strategies in the Entertainment Industry ... 7

1.4 The Dutch and UK Market for Music... 11

1.5 Thesis Outline ... 12

2. Literature Review... 13

2.1 Perceived Value ... 13

2.2 Perceived Value in E-Commerce... 13

2.3 Pricing ... 17

2.4 Bundling... 19

2.5 Versioning... 20

3. Hypotheses and Model Framework ... 22

3.1 Features and Pricing... 22

3.2 Willingness to Pay ... 23

3.3 Features/Extras... 24

3.4 Features versus Price... 25

3.5 Music Involvement ... 26

4. Research Design... 27

4.1 Research Type... 27

4.2 Attribute Selection ... 27

4.3 Choice-Based Conjoint Method... 29

4.4 Data Collection ... 30

5. Results and Analysis ... 33

5.1 Hierarchical-Bayes Analysis... 33

5.2 Willingness to Pay ... 36

5.3 Music Enthusiasts ... 38

5.3 Latent Class Analysis for Segmentation ... 39

5.4 Comparing Approaches: Latent Class versus Hierarchical Bayes... 44

6. Conclusions... 46

7. Limitations and Recommendations... 49

References... 51

Appendix 1: Revenue Numbers for the Dutch Music Market ... 56

Appendix 2: Interview Outline and Summaries... 57

Appendix 3: Summary of the online survey ... 67

Appendix 4: Hierarchical-Bayes Analysis... 72

1. Introduction and Research Questions

1.1 Introduction

The last decade the entertainment industry, be it the music, film, or games industry, has been undergoing radical changes. Changes which made industry participants rethink their business practices and overall management strategies. The changes in these large entertainment industries largely stem from the rise of the internet (Van Eijk, Poort and Rutten, 2010). More and more people are nowadays connected to the internet, with connections that become faster almost every quarter. The United Nations telecommunications agency shows that the past five years the amount of people connected to the internet has doubled, indicating a rapid rate of increase.1

Better connections, more technologically advanced computers and a more optimized structure have led to major changes in the way people share and use information around the world. Amongst this sharing of information, entertainment is also shared. Music, movies, and games can be downloaded on a user’s computer without people leaving the comfort of their homes. People make comfortably use of these possibilities, be it legally or illegally. Major players in the most important home entertainment industries have had long struggles with the changing environments (Van Eijk et.al, 2010). Fighting piracy (the illegal downloading of copyright protected material), and being innovative in the rapidly changing environment, has brought some success stories (i.e. Napster and iTunes) as well as hopelessly failed attempts to reshaping the market2. Today record labels, film distributors and game companies are still experimenting with download services: offering the right value to the customer, and shifts in customer channels from retail to online (Fox, 2004). In this thesis I will focus on the music industry and examine what challenges the industry is still facing. A quick glance at the movie and games industry will provide the music industry new opportunities in their changing market environment. This could provide the music industry with new possibilities to increase market share or improve profitability. Walsch, Mitchell, Frenzel, and Wiedmann. (2003) pointed out quite early that there are possibilities for creating value through ‘premium content’ offerings, which will be a main point in this thesis.

offered to different consumers. Products can be offered in different versions to target different customer segments. Styven (2007) points out that, in a sense, customers actually segment themselves. The customer makes the choice for a certain version. This choice makes clear a customer’s value perceptions and willingness to pay. In a bundling strategy different products are offered, which are combined into one package. Different bundles of products can be created to address different types of consumers, based on their preferences and willingness to pay. In this way the market can also be segmented. In the market for online music, the main (and mostly only) option is to receive the songs as MP3 files, and to pay a certain amount of money for this. In this thesis I will examine whether offering products (apart from the standard MP3 files) that include more than just MP3s has a positive effect on people’s willingness to pay for music. Extra features in a purchase of online music could be discounts to concert tickets, physical media, and physical ‘goodies’, or VIP treatment for customers that are willing to pay substantial amounts of money for their favourite artist. Important aspect to note here is that music is seen as a hedonic experience good. Moe and Fader (2001) state that music is a good driven by the experience provided, and not necessarily by its bundle of attributes. A person’s taste for different music is assumed to play a more important role for the valuation of music, rather than the attributes of the product music. However, this does not mean that construction of different feature-packages for the product of music is assumed to have zero effect on people’s value of music. Including extras can lead to a more positive evaluation of the value contained within a music purchase as well as make people more willing to buy music.

To see what extra features are interesting from a consumer standpoint, I conducted a small amount of interviews. These are used as a preliminary starting point for the quantitative research in this thesis. The conjoint-based quantitative research will be based on theory on perceived value and consumer purchase intention.

The main research question is posed as follows: “Does offering different versions and bundles of recorded music and related products make people value music more and does this increase people’s willingness to pay for online recorded music?” Based on this main research question several other questions are posed:

- What effect does an increase in the number of extra features in recorded music have on the willingness to pay for the product (compared to the standard offering)? - Which extra features of the product have the greatest effect on consumer’s

willingness to pay (compared to the standard offering)?

1.2 Different Business Models in the Online Channel

In the online channel for music distribution several business models are used by different organizations. Fox (2004) specified the major important models. Important here is to note that the records labels mostly do not offer their own services or models. The labels provide music to the different online retailers digitally, which will then be sold to consumers. Pay per Song/Album is the most widely used model in the online environment. Many different sites (i.e. mp3.com, eMusic.com, Amazon.com, and even Walmart.com) on the web offer an extensive library of songs, from which a customer can choose. Customers pay per song (varying from 50 cents to 2 euro per song) or per album (varying from 5 to 20 euro per album) and don’t need a specific account or download software to their computer. Using dedicated software this ‘pay per song’ model is also widely used. iTunes is market leader when it comes to downloading songs and albums. iTunes differs from standard websites because special software has to be downloaded and an account has to be created. With this account, copy-protected and ‘regular’ MP3 files can be combined into one library. Subscription services ask a monthly fee which is paid by customers. For this monthly fee a customer can access music from the library of the service. This can either be unlimited or with a limit on the number of songs and albums. A popular subscription-based service nowadays is Spotify. Spotify entered the Dutch market in 2010 after big successes in for example Sweden and the United Kingdom3. Another example is Napster.com, which started as a peer to peer service offering the sharing of pirated music. Apart from letting people pay a monthly fee for unlimited access to music, Spotify also offers users free-access to their library. In turn, the user will have to listen to advertisements every ten to fifteen minutes. This advertising based strategy therefore also is a possibly successful model. Important to note here is that the large international labels such as Sony BMG and Universal have not had success building their own online distributing platform.4 Apart from this, the major record labels have bundled their resources into creating services similar to iTunes: MusicNet (MediaNet nowadays) and Pressplay. Both services have been unsuccessful, with Pressplay closing after only one year in business.5

1.3 New Strategies in the Entertainment Industry

traditional production schedule of an album contains a major role for the record labels. The labels provide the money an artist can use for creating an album. After distributing the album the label earns back the forward-payment they gave to the artists for developing the album. After this has been paid the lion’s share of the album revenues are further claimed by the record company. Artists receive only a small portion from album sales and thus largely rely on other means to earn money. In this case merchandise sales and revenues from tours provide better income for the artists themselves. The internet provided new possibilities for artists to turn the tables and develop and promote albums themselves. At first this created only a small shift in the hierarchy in the music industry, but the changes might be becoming more and more profound. Some examples are outlined below.

Album for free: Artists who choose to produce their own album (and thus pay for the costs of production themselves) can decide to release the album for whatever price they feel is right. Some artists have decided to provide an album for free, such as Prince and Radiohead 6. Offering an album for free created much media attention, which consequently caused more people to get interested in their music. This eventually led to more sold concert tickets and merchandising (with which the artists earns the most money) (Connolly and Krueger, 2005). Record labels are out of the picture here, because the album is paid for and distribution is organized by the artist (which is made easier using the internet). Marketing and communication efforts are minimized because of the media attention such a move creates. For less well-known artists, this tactic provides a different array of possibilities. Social media such as LastFM and MySpace can provide new artists with the platform to release their music to enthusiasts, without going to the process of production and distribution by a record label. Also, providing music for free using these channels can provide new artists new attention, possibly launching their musical career.

Ten dollar project by software distributor Electronic Arts: In the market for games distributor Electronic Arts has been experimenting with a value-proposition for customers. Electronic Arts has introduced the ‘project 10-dollars’7. This project concerns the increase of the value of the product (in this case, a video game) without changing the price of a video-game. A game (at regular price) is offered with a code, which provides extra content (which can be downloaded from the EA servers) for those who buy new (and legally). People who pirate the game do not have the possibility for downloading the extra content for free. The same holds for people who buy the game second hand. The opportunity to download this extra content is offered regardless, but it will not be free (it

6

http://newsroom.mtv.com/2010/07/07/prince-internet-is-over/ (accessed 02-11-2010)

7

will cost ten dollars). By doing this, people are stimulated to buy a new game. A second game distributor, THQ, supports this notion and is experimenting with a different construct: “When we lower the price to a mass market price the thing really jumps. So what we’re doing this time is we’re coming out initially with a smaller game at a lower price point. We’re then doing a download model for different [extra] [content]. It’s what we call a hybrid - it’s a bit of the micro transaction and [downloadable content] model.”8 Here we can see that the attributes of the product are varied and not the price, which can also have an effect on the way people perceive the value of a product. Combining this idea with the characteristics of a “360-deal” business approach to the music industry, there are new opportunities for varying the contents offered in recorded music.

360 deals: The 360-deal is a new way for artists to push away the record label in favor of more transparent and centrally organized management. Recent mega contracts concerning a 360-deal are Madonna9 and Jay-Z10. Both Madonna and Jay-Z signed with concert-organizer LiveNation (who had no experience with producing and promoting an album) to distribute and promote their new album, organize their world tour, as well as managing supply for merchandising. What makes this interesting is that the clear division between what a record label can offer and what a concert-organizer can offer is fading. This kind of package-deal can provide the artist, as well as the organization, with new ways to combine marketing efforts. Consumers can be offered bundled products, which can increase the value for the product. With or without adjusting the price for the music, the merchandise, or the live tickets, a new value proposition for customers can more easily be constructed.

arguments: 1) the ‘real’ money is made in the smaller market segments where people are enthusiastic about music and willing to pay more money for it; 2) giving up revenue in the larger (low revenue) market segments (for example through offering music at no cost) will potentially increase revenue in the smaller, high revenue, segments.

An example was given by Scott Cohen, who indicated that when people get more interested in something, for example Yoga, their expenses (and willingness to spend) increase rapidly. The same holds for music: people can become more enthusiastic and will spend more money on it. The reasoning is that music should be as accessible as possible, through services but more importantly with accessible prices (free music). With as many people getting access to as much music as possible, the chances of people becoming high revenue enthusiasts are becoming larger. To capitalize on these possibilities both Cohen and Ranasinghe propose the use of different bundles or versions of the music an artist has to offer. As examples they point out including a T-shirt, poster or other articles with an online album purchase. Cohen also explains that there are possibilities for different levels of premium offerings, indicating that there is even a market for people who would like to buy a (signed) guitar with an artist’s new album and spend hundreds of dollars on the purchase.

On top of this both Cohen and Ranasinghe argue that the parties in the music industry should play a different role. When looking at album sales and concert ticket sales (which can easily be bundled), they point out that there is no cooperation between the organizations concerned with these sales. Record labels and concert-organizers are too restricted in their business and should play a different role in the new music environment. Artist-centered management should be preferred, in which everything an artist has to offer is collectively managed. This would create many opportunities for value-increasing bundling efforts, which would have a positive effect on overall revenue. A similar conclusion can be drawn from the previous paragraph, in which the ‘360 deals’ are a result of combining management.

A different approach recently experimented with is a subscription model for physical media. Dutch record label Excelsior launched this subscription model in 201113. For 120 euro per year subscribers receive an album (the CD sent home) every two months. In the months they don’t receive an album subscribers get different items such as scarves, posters, shirts, or other merchandise. Digital media (MP3 or other file formats) is not yet offered, which could be a future extension for the service. The items offered concern only artists connected to the Excelsior label, which gives the label the opportunity to present new and upcoming artists to their fans, as well as bring a new value proposition. Extending on this, it creates a connection between the fans of the artists and the label that

13

is passionate about music. People subscribing to this initiative are labeled “Excelsior-supporters”.

Based on the paragraphs described above one can say that the industry has been trying to adapt to the changing environment. Price has been an important element in these experiments. Artists offering music for free, or for whatever price the consumer wants to pay, as well as record labels and online retailers have tried different pricing tactics to get as many people paying for music has had different results (Bockstedt et al, 2006). Extending on this research, I will use different approaches to pricing strategies to examine what new offerings labels and distributors can bring to the market and be successful.

1.4 The Dutch and UK Market for Music

To see what effects the developments in the music market have had on the sales en revenues of the entire market can be analyzed using the data provided by the NVPI14 (Nederlandse Vereniging van Producenten en Importeurs van beeld- en geluidsdragers), the Dutch market association for the entertainment industry. In this data we can see that sales of physical media (both albums and singles) have been decreasing from 2001, whereas the number of downloads is sharply increasing (Appendix 1). It can also be seen that the strong increase in digital sales is not enough to counteract the drop in sales in physical media (total negative revenue growth of 5% in 2009) Also, the number of high-selling albums also has decreased over the years. Overall it can be concluded that the market for music recordings has been shrinking in the Netherlands. More recent numbers have been released for the UK market from which a decline in overall (physical plus online) sales can be seen15. Geoff Taylor, BPI (the British Recorded Music Industry) Chief Executive, concludes that only since 2010 have online album sales reached the mainstream market. From these numbers it can be distilled that sales of single tracks are most important in the online environment. This is supported by the increasing numbers of singles sales over the past year, whereas album sales (physical plus online) have been declining. Giving consumers the possibility of downloading separate songs from an album has changed the way in which people buy music. This last statement makes the goal of this thesis all the more relevant. Finding out what makes people more willing to pay for music, when given the right product, is therefore the main goal. This can be done through pricing efforts, as well as varying the features included within a music purchases.

1.5 Thesis Outline

This thesis is divided into several chapters, as outlined below: - Chapter 1: Introduction

- Chapter 2: Literature Review

- Chapter 3: Hypotheses and Model Framework - Chapter 4: Research Design

- Chapter 5: Results and Analysis - Chapter 6: Conclusions

2. Literature Review

2.1 Perceived Value

An important question one can ask is “What makes people buy music?” This simple question may necessitate looking at a broader question: “What makes people buy goods and services?” This, of course, is far too broad of a question to answer in this thesis (if there is a clear answer at all), so the focus of this research will be on perceived value and the willingness to pay for good/services (and music in particular). In general, Slater (1997) has stated that “the creation of customer value must be the reason for the firm’s existence and certainly for its success”. With this statement in mind, creating customer value can be seen as one of the most important elements in creating a successful marketing strategy. Although value is deemed as an important element in marketing, there is no clear consensus regarding the definition of value. At the base of the discussions regarding value lies an article by Zeithaml (1988). Zeithaml defined value as follows: “Value is the consumer’s overall assessment of the utility of a product based on perception of what is received and what is given”. This definition shows that consumers make a trade-off between what they get, and what they give up. With this in mind, one can look at the theory of utility, in which consumers derive value from the difference between the utility received by the attributes of a good and the disutility given up by paying for a good. (Tellis and Gaeth, 1990)

2.2 Perceived Value in E-Commerce

Chen and Dubinsky (2003) have developed a framework that points out the key factors that influence customer perceived value and purchase intentions in an online shopping environment. Chen and Dubinsky have specified that the framework is focused on pre-purchase consumer evaluations of a product (and shopping environment). The approach taken in this framework differs significantly from the literature outlined above. Similar to other research, this framework includes price and perceived product quality, but also includes elements specific to a general online shopping environment (valence of experience, E-tailer reputation, and perceived risk). For the purpose of the research in this thesis these elements are assumed to be a constant. However, they are explained shortly.

directly influences both perceived risk and perceived product quality according to Chen and Dubinsky. An online retailer can build reputation through word-of-mouth communication, advertising and investments in brand equity. The consequences of positive E-Tailer reputation are that consumers assume quality persistence and tolerate higher prices for products. Perceived Risk concerns the perceived uncertainty and other consequences when buying a product or service. Chen and Dubinsky point out three types of perceived risk: financial risk (concerning post-purchase service and warranty); performance risk (the quality of the product is lower than expected); and privacy risk (loss of privacy due to collection of financial account data). It is found that perceived risk has a negative effect on perceived value and that it acts as a mediator the relationship between product quality and perceived value.

The framework developed by Chen and Dubinsky (2003) also shows a direct relationship between product price and perceived quality, as well as a direct relationship between product price and quality, and perceived value. These positive relationships seem obvious. A higher price leads to higher perceived quality, and higher perceived quality leads to higher perceived value. The relationship between price and perceived value is a little more ambiguous. It can be argued that, since a higher price implies higher quality, and higher quality implies higher value, that the relationship is (indirectly) positive. However, it can also be argued that a direct negative relationship occurs. When a product is bought, value is created not only through the benefits of the product, but also through what is given up (money).

Figure 1: Purchase-intention framework by Chen and Dubinsky (2003)

has been converging across legal and illegal alternatives, which might suggest that consumers evaluate pirated music more favorably. However, recent research into this specific topic has not been performed. More importantly and relevant to the research in this thesis is the lack of specification of attributes in the model provided by Chen and Dubinsky. To specifically include the notion of attributes into a theoretical model, the framework of Zeithaml (1988) can be used.

Zeithaml’s definition of value is explained using linkages among product attributes; the perceived consequences of consumption; and the personal values of consumers16. Zeithaml described the grey benefit chain (Product
Functional Benefit
Practical Benefit
Emotional Payoff) to link product attributes to perceived customer value. Including the personal values of consumers into the construct of value creates a more psychological approach to a product’s attributes and price. It indicates that consumer’s personal perceptions of price and quality (of attributes) need to be taken into account, given that each consumer has a desired personal payoff. This led Zeithaml to conclude that consumers do not evaluate products on the basis of the objective price and quality, but rather their perception of this price and quality. Quality is therefore substantially different from value. Zeithaml explains that value is more individualistic and personal than quality and therefore value is a higher level attribute. When looking at the framework provided below (Figure 2) it can be seen that Zeithaml (1988) distinguishes between ‘intrinsic attributes’ and ‘abstract dimensions’. Intrinsic attributes are the physical aspects of a product, such as packaging, contents, materials used et cetera. The abstract dimensions are explained as the final step in the benefit chain pointed out earlier: the emotional payoff attained when using a purchased product. A distinction between objective (actual) price and perceived (monetary) price of a product is also made. Zeithaml argues that the distinction is made because there are large differences in the way consumers ‘encode’ (evaluate) prices. Some people ‘encode’ prices as they are, and others just use definitions as ‘expensive’ or ‘cheap’. Some don’t ‘encode’ prices at all, which makes a concept of perceived prices all the more important. The other elements in the framework do not have direct relevance for the research in this thesis, so no detailed description is needed.

16

Figure 2: Perceived-Quality framework by Zeithaml (1988)

Khalifa (2004) uses a different approach in defining perceived value and uses three different categories of perceived value.: ‘customer value in exchange’ (the tradeoff in receiving and giving up); ‘customer value build-up’ (focusing on the benefits in the equation); and ‘customer value dynamics’ (how consumers evaluate a product or service

over time). These three different categories divide several stages in the valuation-process of a consumer. One way in which the division is made is based on the stage in the buying process (for example the third category concerns only the valuation before buying. Woodall (2003) created a more extensive division in the different purchasing-stages in which people can value a good or service: ‘ex-ante’, ‘transaction’, ‘ex-post’, and ‘disposal’. Research in the field of value makes it clear that value is a complex element in marketing.

described the grey benefit chain such a chain could also be constructed for the emotional elements in a product such as music. A chain could look like this: Product
Emotional response
Emotional Payoff. Including this emotional construct into the model provided by Zeithaml (1988) provides a solid basis for the framework used in this thesis.

2.3 Pricing

When looking at a product’s price two statements can be made. Firstly, there is a direct and positive relation between product price and product quality. Secondly, there is a direct and negative relation between product price and a product’s ‘value for money’ (Zeithaml, 1988). Zeithaml concluded that the negative effect on ‘value for money’ is most important here. The definition of ‘perceived price’ arose in Zeithaml’s research to include a concept for quality in the definition of price. Chen and Dubinsky (2003) show that, for an online shopping environment, price becomes an important element in consumer’s valuation of a product. They base this conclusion partly on research by Korgaonkar and Wolin (1999) who stated that online shoppers mostly look for the lowest price among different online store alternatives. Price sensitivity has also been proven to exist in the market for online music (e.g. Walsch et al, 2003). On the other hand Lynch and Ariely (2000) point out that, when exclusive content is offered, price sensitivity decreases strongly.

willingness to pay for subscription services depends on the contents of the subscription (i.e. the number of songs available, no limits in validity). This is supported by Walsh et al. (2003), who found that respondents prefer a subscription service over the purchase of individual songs as long as the service offers unlimited downloading of songs. They conclude that people expect a subscription service to be, above all, useful and beneficial. People expect the pay-per-song service to be, above all, transparent in their cost-structure. Wunsch-Vincent and Vickery (2005) point out that changing consumer preference in the market for online music are also changing the way pricing strategies need to be applied. They point out two important developments in consumer preferences that should, in their opinion, lead to new pricing strategies. First is the way people consume music. According to Wunsch-Vincent and Vickery music is more broadly consumed, which means that not only music, but also artist related news; social events (be it online or offline); and lifestyle are important. Secondly, they point out that ‘unbundling’ can be detrimental to the music industry. People now have the opportunity to download separate songs (which they like) of an album instead of having to buy the entire album. These findings are supported by recent numbers from the UK music industry (as detailed before).

Styven (2007) states that, based on the above outlined research, customers are strongly aware of the value that is offered through online channels and have many alternatives in finding their optimal providers. Firms face high fixed costs, but (due to the online retail channels) greatly decreased variable costs. Given the customer’s preference and willingness to pay and the current cost structure for firms, a value based pricing strategy has to be adopted. Styven offers different strategies in this field for online music services. The notion that strategies are changing is supported by Leonard and Kusek (2005), who state that current strategies will be replaced by a ‘liquid pricing system that incorporates subscription, bundles of various media types, multi-access deals, and added value services”. However, it can be stated that the willingness to pay for individual customers is hard to determine. Inoue, Nakajima, and Yoshikawa (2001) point out the possibility of creating a generalized form of willingness to pay through the use of data on information exchange and purchase behavior.

also raise concerns to this strategy. They point out that privacy can become a significant issue when more and more information about the customer is stored and used for marketing strategy. They also state that customers may feel that prices are unfair. When customers find out that prices differ on an individual level they can feel they are not treated fairly and brand loyalty can be damaged. The concept of intertemporal price

discrimination was explained to be potentially successful by Gallaway and Kinnear (2001). They found that customers are willing to pay more for new music compared to older music. Walsch et al. (2003) state that targeting different segments of customers and differentiating music offerings based on recency or popularity may be a successful strategy.

Windowing is a strategy used in the film-industry in which movies are released through different channels through a time-schedule. Earlier viewings of a movie go through certain channels, which offer a different experience, but cost more. People thus pay more to 1) see the movie immediately when it is released and 2) to see a movie in ‘ideal viewing conditions’. Styven (2007) points out that windowing can be seen as a combination of versioning and intertemporal price discrimination. Bhatia, Gay, and Honey(2003) created an outline for a windowing strategy in the market for music. First, an album would be released through the traditional retail channel at retail price. After this, the album would release in the online channel, where individual songs can also be purchased. In the next ‘window’ the album is released on subscription services. Finally, the album can be obtained through budget channels. However, in practice this strategy is applied differently, because the online channels such as iTunes (due to large market share) or Spotify (for communication purposes) get the advantage over the traditional retail channel. It also points out a paradox in the music industry as record labels have more power and potential margin through traditional retail channels, but need the online channels (in which they have far less power and margin) for market-share and communication advantages.

2.4 Bundling

promotions, whereas product bundling can be used to created added value and segment the market. A result of combining separate products is that customer demand is optimized at one single price per bundle (Inoue et al. 2003), and customer valuation becomes more homogeneous and predictable. (Zhu and MacQuarrie, 2003). Eppen and Hanson (1991) bring forth additional benefits. According to them bundling can decrease production and distribution costs, create complementarities, and segment consumers based on their product valuations. This last point is explained in more detail by Zhu and MacQuarrie: bundling can reduce consumer diversity, which makes it possible for producers to extract more consumer surplus in the market. This, in return, would lead to higher profits. Bakos and Brynjolfsson (1999) state that consumer diversity is decreased due to difference in the variance of the probability distribution of individual goods and bundled goods. Bundles of goods have probability distributions with lower variance than those of individual goods. This additional variance leads to more extreme values, and therefore uncertainty, which makes pricing strategies harder to implement. Schmalensee (1984) has shown that a reduction in buyer diversity leads to higher profits among all consumers. Offering different bundles would effectively address different consumers with low valuation, as well as consumers with high valuation. In the online environment bundling can easily be applied, since costs of production and distribution are low compared to physical production and distribution. Stremersch and Tellis (2002) provides three ‘forms’ of bundling that can be used in the market space: with unbundling products are all offered separately; with pure bundling products are only offered in a bundle; and with mixed

bundling both the bundle and the separate products are offered. The mixed bundling approach is used for the research in this thesis (see section 4.2).

2.5 Versioning

appropriate categories.” The customer makes the choice for a certain version, and this choice makes clear a customer’s value perceptions and willingness to pay. Ancarani (2002) points out that this strategy allows for examining customer price sensitivities to find the ‘sweet spot’ for different prices.

Many opportunities for including other products within the music purchase arise here. Walsch et al. (2003) state that online channels in the market for music have “high potential for cross-selling products such as T-Shirts, books, and posters, as well as products required to be effectively download and consume music which could be sold (e.g. computer memory chips, loud speakers, MP3 Players, CD-Rs)”. Shapiro and Varian (1998) indicate that, for an information product (in their example books and movies) the costs of investing are high, but the costs of production and distribution are low. In online industries this becomes a stronger argument, since costs of distributing are minimal and there are virtually no restrictions to shelf-space. With these low costs in mind there are possibilities to address the low end, as well as the high-end of the market using different versions of a product. It is stated that, for versioning to be successful, the most important concept is to determine the right features to people that value these features most. In the best situation, this makes each person pay the highest possible price for the product they value. Furthermore, it is indicated that there can be different successful ‘versions of versioning’: convenience, comprehensiveness, manipulation, community, annoyance, speed, data processing, user interface, image/file resolution/quality and support. Some of these ‘versions of versioning’ can also be applied in the market for recorded music. More on this is found in the next chapter.

Varian and Shapiro even point out the possibility of creating a free version to e.g. new customers. Positive effects of this are summarized as: Gaining follow-on sales, creating awareness, creating a network, and ultimately gaining a competitive advantage. The first positive effect is important for the topic in this thesis. Gaining follow-on sales indicates that people can gain a certain level of interest for the products offered. It can ultimately lead to customers who pay increasing amounts of money for additional products. This is in line with the concept of versioning, as well as the insights of Cohen11 and Ranasinghe12 (see footnotes on page 9).

3. Hypotheses and Model Framework

In the previous chapter an extensive overview of previous research was provided. In this chapter the theories will be used to state several hypotheses as well as a small framework. The hypotheses formulated in this chapter will be used in the data analysis in the following chapter.

3.1 Features and Pricing

The main goal of this thesis is to provide new insights into the way people value music as a product and what makes them more willing to pay for music. Fox (2004) states: “Consumers will be willing to pay for such music only if they believe that they derive real value from doing so. In particular, the record labels will need to provide value over and above that provided by free music”. It was previously discussed that an increase in attributes or features could provide for many opportunities in the market for online music. When looking at research into attribute composition for ‘regular’ goods Tversky and Kahneman (1991) have pointed out a ‘reference-point’ model. In this model it is assumed that consumers make reference points regarding each of the individual attributes of a certain good. With this model Tversky and Kahneman bring forth three conclusions that are used for constructing the hypotheses in this thesis: 1) a gain in a product’s attribute, compared to the consumer’s reference level for that attribute, affects a product’s perceived value positively; 2) a loss in a product’s attribute, compared to the consumer’s reference level for that attribute, affects a product’s perceived value more strongly than a gain in that attribute; 3) gains or losses in attributes, compared to the consumer’s reference level for the attributes, have diminishing effects (the first gains or losses have the largest effect) on a product’s perceived value. It makes it important to state a reference level for an online music purchase. Since almost all online music transactions are the same, a reference level is easily stated. Currently most transactions are standard quality MP3s for a price of €9.99. This reference level will be used more in chapter 5. As noted in the research question, the ‘attributes’ or features contained within the product of recorded music can vary. Based on the ‘360-deals’ and ‘Ten dollar project’ detailed above possible extensions to regular online music (album) purchases can be outlined as follows:

- Offering the possibility of downloading higher quality Audio-formats

- Adding exclusive extra (bonus) songs to customers who purchase an album through the online service

- Adding exclusive extra products to an album, such as a t-shirt or poster, to people who purchase an album through the online service.

- Adding a discount to a future purchase through the online service (music or other products)

- Adding a discount to tickets for a concert of the artist after an album purchase through the online service.

Combining the frameworks developed by Chen and Dubinsky (2003) and Zeithaml (1988); and the pricing theory summarized by Styven (2007) a framework can be constructed, based on the following elements:

3.2 Willingness to Pay

The OECD17 defines willingness to pay as follows: ‘The stated price that an individual would accept to pay for avoiding the loss or the diminution of an environmental service’. People state what they are willing to pay, but not how much they actually do pay in a certain transaction. Related to the willingness to pay is the willingness to accept18: “The stated price that an individual would accept in compensation for the loss or the diminution of an environmental service”. The willingness to accept therefore looks at the selling party in a transaction. When looking at the relation between product value and the willingness to pay for that specific product an immediate glance at microeconomic theory can be taken. In microeconomic theory we find the concept of utility maximization: the tradeoff between what we get and what we give up is most important here. If what we pay is constant, but what we receive has more (perceived) value, the tradeoff becomes more interesting for us as a consumer. Based upon this, one can argue that when one receives more (value), one is also willing to give up more in the tradeoff between money and the product. This tradeoff is also found in the definition of perceived value by Zeithaml (1988): ‘Perceived value is the consumer’s overall assessment of the utility of a product based on perceptions of what is received and what is given’. Chen and Dubinsky (2003) have stated that the direct (positive) relationship between perceived value and the willingness to pay for music is indeed present in the online purchasing environment. These results contribute to previous research (Monroe, 1990; and Rao and Monroe, 1989) in which this relationship was proven in a traditional brick-and-mortar channels. In the market for online music this relationship is assumed to be present as well, leading to the following hypothesis:

H1: Perceived value is positively related to the willingness to pay for online music.

17

http://stats.oecd.org/glossary/detail.asp?ID=2929 (accessed 03-04-2011)

18

3.3 Features/Extras

With the conclusions provided by Tversky and Kahneman (1991) before, it can be stated that the proposed increases in features or extras for online music purchases have a positive effect on consumer willingness to pay. Brown and Carpenter (2000) also state that every added feature is yet another reason for consumers to buy a certain product, but they also state the effect of trivial attributes on choice. Trivial attributes are attributes that are stated as being irrelevant for a product purchase. However, as Brown and Carpenter (2000) state, these trivial attributes do have an effect on people’s product evaluations and decision making. Trivial attributes can have both a positive and negative effect, based on the way products or brands are differentiated from another.

Simonson, Carmon, and O’Curry (1994) state that bundling a product with irrelevant promotional offers can reduce brand choice. They have stated that addition of unneeded features or extras can have a negative effect on people’s perception of value and quality, when the price of the package is not increased. Consumers can think that they are paying money for an unneeded feature, and therefore reduce the perceived value of a that particular base-product. A different negative effect of an increase in the number of unneeded features can be pointed out: people can have the feeling that their choice was influenced by this irrelevant feature. This can be inconsistent with a person’s self-concept, according to Simonson et. al (1994). People can for example indicate that they are ‘not the type for collector’s editions’ because they have certain (negative) associations with collecting items. In addition to the conclusions by Simonson et al. (1994) Meyvis and Janiszewski (2002) indicate that consumers primarily look for product information about desired benefits. Other information, be it irrelevant or not supporting their desired benefits, is all treated as being not-supporting the belief that a desired benefit is included. This would indicate that irrelevant or unneeded benefits and the information on these benefits can weaken a consumer’s belief that a desired benefit is indeed also present. These conclusions indicate that the inclusion of certain features and extras should be done with great care.

Thompson, Hamilton and Rust (2005) state that, from the perspective of the selling-party, adding too many features can have a negative effect on customer lifetime value and future sales. They indicate that having a product that includes a number of features that maximizes initial choice can have a negative effect on consumer satisfaction and therefore future sales. For this reason they recommend bundling and versioning strategies: having different specialized products with a limited number of features is preferred over a one-product offering which includes as many features or extras as possible.

different from another, apart from the musical composition. Styven (2007) has pointed out that sound-quality and quality of the service are important attributes for online music consumers, which is a useful basis for the research in this thesis. It can be noted that the different feature increases are assumed to have substantially different effects. A final remark to make is that the extras offered are in accordance to what consumers point out to be as ‘good’ extras. For example extras that exploit the artist or bring forth ethical concerns are not defined as ‘good’ extras. More details on the specific extras can be found in section 4.2.

H2: The number of extras contained in recorded music has a positive effect on the

willingness to pay for recorded music.

3.4 Features versus Price

It is already assumed that there is a positive relation between the number of attributes and the perceived value of music. Also the negative relation between price and perceived value has also been outlined. It would be interesting to see which effect is stronger. It is proven that price is an important factor for people when buying music online: Korgaonkar and Wolin (1999) argued that online shoppers mostly look for the lowest price among different online store alternatives. Walsch et. al (2003) have concluded that this also holds for online purchases of recorded music. Therefore it is assumed that increasing the number of features accompanying the purchase of music has a smaller effect on perceived value (and willingness to pay) than lowering the price of the music purchase. Furthermore, as explained in the previous section, the relevance of the added extras is important. Extra features that are seen as irrelevant are likely to have a small effect on a consumer’s willingness to pay. Even negative effects are possible (Brown and Carpenter, 2000 and Thompson et.al., 2005). On the other hand the conclusions by Lynch and Ariely (2000) would indicate that, when the number of features added to a product leads to greater exclusivity of this product, the negative effect of price could be overstated in this particular case. On the other hand Thompson et al. (2005) argue the opposite, as seen in the previous section. This makes including the right extras all the more relevant for the research in this thesis. The process of selection of the ‘right’ extra features is outlined in chapter 4.2. Important to note is that, although price is a fundamentally different attribute than any other product attribute, the effects on perceived value can be analyzed through individual utilities. More on this can be found in chapter 5. This brings forth the following hypothesis:

H3: The positive relation between the number of features and perceived value is stronger

3.5 Music Involvement

Music involvement or enthusiasm also plays a role in the process of purchasing music (Walsch et al., 2003). However, as Styven (2007) stated, no research into this specific field has been made. Styven therefore suggests the use of the following definition for music involvement/enthusiasm: ‘the extent to which a person expresses an interest in a

product category and views it as an expression of his or her lifestyle’. According to Styven (2003) the concept of involvement can be measured through the usage rate of online retail channels such as iTunes or sites from which music can be bought. Information on usage of these channels can easily be obtained through surveys, which makes this an interesting point of research. Walsch et al. (2003) propose a somewhat similar measure, but specify this to ‘music-sites’. People visiting music sites regularly (be it to look for information or to download/listen to music) are counted as ‘intensive’ music consumers. Based on these thoughts the following hypothesis can be put forward:

H4: (Online) Music enthusiasts have higher willingness to pay for online music than

other (online) music purchasers.

Combining the theory developed by Chen and Dubinsky (2003) and Zeithaml (1988) with the pricing theory summarized by Styven (2007) the following framework was constructed to answer the research questions posed above. Note that demographic variables (age, income, online music purchaser yes/no) are used as control variables.

4. Research Design

In the previous chapter the research hypotheses and conceptual framework were developed. In this chapter the outline for the quantitative research will be provided. This chapter will provide information on the type of research, the data that was collected, the survey that was constructed, and the statistical tests that have been performed.

4.1 Research Type

The research in this thesis has an experimental research design. In an experiment a causal relation between an independent variable and a dependent variable is analyzed (Baarda & de Goede, 2000). The dependent variable in this case is the intention to buy online music. This thesis mainly concerns the online purchases of full-length albums. There are two important reasons for this: 1) as identified by the BPI12, full-length album sales are lagging behind when it comes to online music purchases; 2) the price (and margin) concerning a purchase of an album allow for more experimenting with features. In this thesis consumers thus make an evaluation when looking at different sets of attributes. For this specific type of research a conjoint analysis is most appropriate. Malhotra et al. (2007) give the following definition: “Conjoint analysis attempts to determine the relative importance consumers attach to salient attributes and the utilities they attach to the levels of attributes”. As in any research design based on this type of analysis respondents are asked to evaluate different sets of attributes contained within the online purchase of an album. From these evaluations it can be distilled which attribute is most important for consumers. Malhotra advises to limit the amount of attributes to six or seven, because respondents otherwise would have a difficult task of evaluating the alternatives. However, when looking at different attribute levels for these attributes six or seven attributes seem to be too much. At the start of the chapter I have proposed seven different ways to ‘upgrade’ the purchase of an album. Some of these upgrades are similar to others. This requires some narrowing down, which ultimately makes the task for respondents easier. To achieve this, a small qualitative research-design was made. Interviews were held to find the added features that first come to mind with consumers. More on this is in the next section. The results of the interview were used in constructing the quantitative research. Section 4.3 and 4.4 will discuss this.

4.2 Attribute Selection

may play a role, they are very hard to make stand out in the online channel, since there is no physical transaction. In previous research (Dufft et al., 2005; Walsch et al., 2003; and Gallaway and Kinnear, 2001) it was shown that price is an important attribute. Research has found fairly similar results when looking for the ‘sweet spot’ in pricing for music for both singles (Walsh et al (2003) and others) and albums (Lee and Gosain, 2002). Based on this research price will be included as an attribute (with different levels).

Brand name is also a regular attribute in research using conjoint analysis. However, when looking at music the definition of ‘brand’ is difficult to pinpoint. Record labels are a type of brand, names such as Motown or Blue Note can be critical to an artists’ success. The artist’s name is the most important concept of a brand when it comes to music (Bockstedt et al, 2006). Since there are many artists in many different genres it is nearly impossible to make a brand attribute when it comes to conjoint analysis. People all have different artists that all evoke different emotions and possible purchase intentions. Respondents do need to be reminded of an artist who they like when participating in the choice based survey, since a disliked artist would not evoke any purchase intentions regardless of extra features or lower prices offered. One way to make sure respondents think of an artist they have positive feelings about is asking them about their favorite artists at the moment. Respondents will be asked to fill in their favorite artist so they are likely to think about this artist as relevant for the choice tasks.

- File format considers the files that are downloaded to the hard drive after purchasing online. Normal MP3 refers to 192kbps (bit rate) MP3 files. HQ MP3 refers to 320kbps (bit rate) files. WAV and FLAC are lossless codecs, which have CD quality sound.

- Physical media concerns the types of CDs people want to receive by mail. The regular edition is a normal edition with no extra songs. The special edition has extra songs and a special cover. A live DVD is a concert registration on DVD. - Online community is about the access people have to an exclusive online

platform. When buying an album online people can receive an access code with which they receive extra information and media. Limited access means that there is access with extra information and b-sides. Full access also includes live footage of the artist (and others).

- Price speaks for itself. A large number of levels were chosen to simulate different market situations (prices are above/below certain current market prices for albums or CDs/DVDs).

Table 1: Attributes and their levels, used for choice-based conjoint study Attributes Levels

File format Normal MP3 HQ MP3 or FLAC/WAV

Physical media No CD/DVD Regular edition CD Special Edition CD Live DVD Online community No access Limited access Full access

Price 7,99 9,99 14,99 19,99 29,99

4.3 Choice-Based Conjoint Method

Before acquiring actual results from surveys a conjoint analysis method needs to be determined. The choice of method will determine how the surveys are constructed. Orme (2009) points out that there are several different techniques in conjoint analysis that are suitable for different situations and samples. For the purpose of this thesis it is not necessary to describe the pros and cons of every method, but it is useful to look at some different techniques that are useful in this specific type of research. In a

Choice-Based-Conjoint survey different sets of attributes are given. Respondents need to choose the alternative which suits them best. This type of survey is easy for respondents and it delivers data that is easily analyzed. It is also recommended for study in which few attributes are analyzed, and interactions are assumed to occur. On the other hand, there may be a problem with reliability. This is because some respondents could get bored or fatigued from repeatedly performing a similar task. A different approach is the

Elrod, Louviere, and Davey (1992) examined the differences between the two different approaches and found mainly advantages for the choice-based approach. Firstly, because choice is ultimately what marketers are interested in, the choice based approach should be more suitable, according to them. Furthermore the values and statistical significance of the parameters included in the analysis are more easily reported. On the other hand there is a possible issue when looking at smaller sample sizes: ‘special’ tastes of some respondents (or faulty responses) can have a strong influence on the calculations based on the respondent’s choices. According to Elrod et al (1992) this can be overcome by using aggregate level data. This, however, would lead to limitations regarding predicting consumer choice on individual and segment-specific level. A rating-based model is able to provide this information more easily. On the other hand it is stated that estimation and statistical testing and aggregation are more difficult and cumbersome when applied. These conclusions are in line with the findings of Moore (2004), and Moore, Gray-Lee, and Louviere (1998), who indicate that the Hierarchical Bayes method used in choice-based conjoint analysis has optimal performance compared to ratings-choice-based, OLS, and latent-class analysis. Their conclusions are based on hit-rates and choice share validations.

Apart from this there is the consideration of the data collection method. Rao and Hauser (2002) point out several methods in which respondents are required to make choices. A consequence of this is that there are several ways for data collection. Partial profile evaluations let respondents evaluate products based on comparing individual attributes to determine which attribute has the most effect. However, the entire product is not fully shown this way, which could cause the analysis to exclude any kind of interaction between attributes. The full-profile evaluations do include different sets of attributes and consider the product as a whole, therefore making it possible to analyze possible interaction effects. With this approach choice-based data can also be generated. Rao and Hauser (2002) describe this as ‘stated preferences’ because respondents state a preference for one product over the others in the choice task.

Based on these techniques it would be most appropriate to use the ‘traditional’

Choice-Based-Conjoint survey as a basis for the quantitative research in this thesis. This leads to the creation of several product profiles (with sets of features or extras) that respondents need to choose between in different sets.

4.4 Data Collection

of the survey is clear and the results are ready to be used in statistics software. Apart from this the online survey method takes the least amount of time, which is important since a choice-based survey can become repetitive or dull rather quickly. Furthermore it gave me the opportunity to reach a larger sample, which improved the quality and reliability of the analysis. Respondents were each given twenty choice tasks. Including too many choice tasks would to a decrease in data quality due to respondents getting bored or fatigued. It was recommended that fifteen to a maximum of thirty choice tasks are presented to the respondents (Orme, 2009 and Oppewal, 1995). It can be stated that twenty choice-tasks is still a high number of tasks for respondents to complete. However, being an online survey made sure filling in the survey was a speedy and easy process for respondents, which justifies the choice of a somewhat high amount of choice-tasks.

I have used the Sawtooth Software SSI Web tool for creating a conjoint study. This software makes it possible to fill in the attributes and their levels and create an orthogonal design based on different methods. I chose to use the ‘balanced overlap’ approach for reducing the number of profiles that have to be compared by respondents. This method was chosen for different reasons. First there is the variation in the number of levels per attribute in this analysis (some include three or less levels). Using the traditional approach of minimal overlap between attributes’ levels would simply be impossible; therefore the minimal overlap approach would be compromised. According to Orme (2009), there are some additional advantages: people who consistently make a choice based on one attribute (and one level) are forced to think when this attribute (and its specific level) are not included in one question. The type of questionnaire structure in the balanced overlap approach included cases where some attribute levels are not included, which could prevent this specific situation from happening, which in return makes product-choice analysis more accurate. It is for this reason that Sawtooth Software also recommends (Orme, 2009) a balanced overlap approach for creating an orthogonal research design.

In creating these twenty choice-tasks no exceptions were made to exclude “extreme” profiles in some of the choice tasks. For example a completely stripped product (only the MP3 files) at a high price was compared to a complete product (MP3 files, Special Edition CD and community access) at a low price. The choice to include these ‘extreme’ options was based on the conclusions by Green and Srinivasan (1990), who recommend the use of these extreme or ‘unacceptable levels’ to ensure the rejection of these possibilities. Currently the large price-variations in the market for (online) music provide a second reason for including extremities in the choice-based survey.

5. Results and Analysis

In the previous chapter it was explained how the data for the quantitative research for this thesis was acquired. In this chapter the data on all 134 included respondents is analyzed using two methods. The first method used is the hierarchical-Bayes method, which uses specific data on all 134 individual responses. The second method uses the individual data to create a number of classes (segments) that can be used for analysis. Both methods can provide valuable insights for pricing strategies, as well as provide information on which features to be included in future product offerings. The hierarchical-Bayes method provides these results on an individual respondent level, which can provide specific data on consumer preferences. The latent class approach offers a number of classes (segments) that can identify interesting segments in the market. Both methods provide marketing managers with information that can be used in creating marketing strategies to target new market segments, or to communicate differently with existing segments. As explained in the previous chapters it was pointed out that the music industry does not make extensive use of these types of strategies. Instead music distributors mainly choose to offer a single product across a large group of consumers that have different preferences. In this chapter these different preferences are distilled using the two analytic methods. In section 5.4 the hierachical-Bayes and latent class methods are compared.

5.1 Hierarchical-Bayes Analysis

Table 2: Respondent characteristics

Average age (min-max) 25 (16-59)

%Male/%female 76%-24%

% of respondents with higher education (median\modus) 78% (HBO) % of respondents with high music interests 72% Median number of hours per week listening to music 15-20 hours Average number of concerts visited per year (std.dev) 4 (4.41) Average number of MP3’s downloaded per month (legal and ‘illegal’ or free) (std. dev.) 49 (49.3) Average number of CD’s bought per year (std. dev.) 4 (6.11) Average amount of money spent on music per month (std.dev.) €20.13 (€51.11) With the Sawtooth CBC module a hierarchical Bayes analysis was performed, which created individual ‘utilities’ for each attribute and their respective levels. The different variables were input as part-worth variables. This included the price variable (which is not fully linear due to an uneven spread between price-levels). Of the twenty choice tasks included in the survey, three were excluded to be of use as out sets. With these hold-out sets the predictive strength of the results were tested. With the data of 134 respondents an analysis was performed using 15000 (total 25000) initial iterations, which created a set of individual part-worth utilities for each respondent. Further analysis, using these utilities, was performed using Excel.

Before performing this analysis, a quality evaluation of the output generated by Sawtooth is required. From the output the goodness of fit measures are important. The average percentage certainty for this particular analysis was 0.739, or 73.9%. This indicates that the log likelihood for this analysis is 73.9% in the way of the value expected by chance (0%) and a perfect prediction (100%). 73.9% can therefore be seen as a good result concerning it is far closer to a perfect prediction than to expectation by chance. This conclusion can also be drawn from the root likelihood value provided by Sawtooth. Here one can see an average value of 0.751. A value of 1 for the root likelihood would indicate a perfect prediction. Prediction by chance would lead to the following attributes (dependent on the different levels): File Format: 0.5; Physical media: 0.25; Online community: 0.33; and Price: 0.2. The average root likelihood is above the expected root likelihood for each attribute. Based on this one can also conclude that this analysis is substantially better than prediction based on chance. Using a total of 25000 iterations, evidence was found for stable convergence since no large differences between the average and last likelihood values could be detected.