Case study Last.fm

Last.fm is an Internet radio station and music recommendation system that records what you listen to, and then presents you with an array of suggestions based upon your taste — artists you might like, users with similar taste, personalised radio streams, charts, and much more (Source: FAQ). Last.fm is a company that employs 16 people, and started from the beginning as two different projects, Last.fm and AudioScrobbler. Last.fm merged with sister site AudioScrobbler in August 2005. The system builds a personalized detailed profile of each user’s musical taste, showing their favourite artists and songs on a customizable profile on the website, comprising the songs played on its stations selected by means of a collaborative filter, or optionally, recorded by a Last.fm plug-in installed into its users’ music player.

Last.fm started in late 2002, and it developed out of a platform for artists and bands. It consisted of a platform where people could upload their own compositions, and of a radio stream on these compositions. The problem Last.fm faced was that nobody knew any of the artists, so the challenge was to present the music in such a way that the right music would be connected to the right listeners, by presenting it in genres or playlists. They realized that the more music they would get, it would be more difficult to fit it in genres, also because putting music into genres is somewhat subjective. The challenge Last.fm was facing was how to find an audience for the music, without having to go to the music press. They thought there was an easier way to connect people with the right music.

AudioScrobbler started as a final year project at the Southampton University. The project consisted of some plug-ins that could be installed on media players like Winamp, iTunes and Windows Media Player. These plug-ins submitted every song that was listened to on these players to a central database, to a personal online music profile. This online music profile consisted of favourite artists, favourite songs, and weekly charts.

In August 2005, Last.fm and AudioScrobbler joined into one site. The basis of the site is the personal music profile. Once a music profile is available, the system starts comparing it with other profiles to find musical neighbours with similar music taste, so new music can be recommended. All other listeners determine the recommendations by the music they listen to, therefore it is a social recommendation service. It is not based on editorial input, or algorithms extracting some properties out of sound files. Each music profile belongs to one person, and describes their taste in music. Last.fm uses these music profiles to make personalized recommendations, match consumers up with people who like similar music, and generate custom radio stations for each person.

Last.fm is included in this research because it matches the site selection criteria of Chapter 3. One of its objectives is to promote music and find the right audience for it, and another is to have as much digital products available to offer. Last.fm is completely community driven (Source: OB). With this community driven approach Last.fm tries to let consumers discover new music from its broad catalogue.

Data sources

The primary data source of the Last.fm case study is an interview I conducted with one of the co-founders of Last.fm, Martin Stiksel. The duration of the interview was just over an hour, and was held during a Skype session which was recorded with permission. The recorded conversation was literally transcribed afterwards, and sent back to the interviewee to review it. Other data sources are documentation in the form of the extensive website of Last.fm, and archival records in the form of some published interviews on the Internet. A complete list with references can be found at the end of the chapter.

Mass customization classification

Last.fm can be classified as an involver in terms of the mass customization configuration, see Table 4.1. The earliest point of consumer involvement is in the design and fabrication stages, while the type of modularity employed is present in the assembly and use stages. The point of consumer involvement is primarily determined by the fact that consumers can specify new product features, consumers’ requests are uniquely designed into the finished product, and each consumer order is a unique design. The type of modularity employed is primarily determined by the fact that options can be added to a standard product, components are shared across products, and products are designed around a common core technology.

 

Group

Modularity

Consumer involvement

Design / fabrication

Assembly / use

Design / fabrication

Assembly / use

1. Fabricators

+

±

+

±

2. Involvers

+

+

±

3. Modularizers

+

±

+

4. Assemblers

+

+

 

Table 4.1: Classification of the mass customization configuration.

Consumer involvement
The earliest point that consumers of Last.fm are involved is during the design and fabrication stages. Consumers can specify new product features, consumer’s requests are uniquely designed into the finished product, and each consumer order requires a unique design. In the remaining stages of the production cycle, the consumers are still involved, because it is possible for consumers to select features from listings. This section gives proof of the consumers’ involvement for these factors.

The AudioScrobbler plug-in that Last.fm users can use on their favourite music player automatically submits the song information from the song that they are listening to Last.fm, and this song gets its own page on the Last.fm website if it does not exist yet. The submitted metadata of new music is immediately available for everyone. Another consumer group is the record label industry, which can also specify new product features by uploading new music.

“There is a section at the bottom of the Last.fm website where you click on a link which is called ‘Labels & Artists’. As a record label you can sign up to Last.fm and upload your music into the radio and if your music is totally unknown, nobody has heard it before, you can give it starting points in the Last.fm system, so for example you can tag it (BR: tags are keywords or labels (Source: FAQ)), you can say that this is like rock and roll, or this is hip-hop or something like this so it has a starting point. Then you can also type in similar artists, so you can say like my band sounds like Legoweld and Kraftwerk and other punk records from Holland. […] When a user is listening to this for the first time, then we already have some information to start recommending it to other people.”

Both listeners and producers are involved early in the design and fabrication stages, by specifying new product features. About 8,000 producers and record companies, small to medium large, work together with Last.fm, and they are uploading music regularly. The music appears on the radio straight away, and it can be accompanied with a cover image and a link to their favourite online store. On the Last.fm website they can have an artist page, an album page and a label page to customize as well and write about the artist or the track.

At some level it is possible for consumers to uniquely design their requests into the finished product. On the Last.fm website there is the possibility to customize the radio with a popularity filter. This filter is a slider that can be set from obscure to popular, which directly influences the recommendation radio station of the listener. Recommended music that can be listened to comes from musical suggestions from users, groups and from the Last.fm platform (Source: DM). Users are other listeners that can directly recommend music to other listeners. Listeners can also join groups which are created by other listeners, and can recommend music to all members of that group. A group can be based on similarities between users, such as their country or a favourite artist, or something else users have in common. Every Last.fm consumer is free to start a group about anything.

Every product, or radio station, is different from every other available product. When a new person starts using the Last.fm radio player, his or her musical profile immediately starts to personalize.

“You have to pick a station to start with, so when you install the player for example it pops up with a window which tells you to type in a similar artist so you can start of a similar artist station which is one way of starting to fill up your profile. So you just start off a radio, there is tag radio stations as well that you can start, but obviously if we don’t know anything about you at the very beginning, we can’t really make you any personalized suggestions yet. Once that we know stuff about you it gets better and better. The longer you use it, the more we know about you the better we can recommend you new things. At the beginning you have to give it a kick, you have to sort of kick it off yourself by typing in either an artist or a tag and pressing play.”

As mentioned, when consumers are involved early in the design and fabrication stages, they are also involved in the assembly and use stages. It is possible for consumers to select features from listings, as is the case when consumers are involved in the assembly and use stages.

“If you click on music (BR: on the Last.fm website) and then on tags, popular tags in the middle, like ‘electronic’, then you see all this things tagged with electronic. And on the right hand side there is a playlist with songs which are all full length tracks, so these are all coming from labels so there is a special playlist for the electronic tag that you can listen to straight away and all these songs are available for free.”

Table 4.2 gives an overview of all the consumer involvement factors of Duray et al. (2000) that were encountered at Last.fm.

 

Consumer involvement factor

Stage in production cycle

Consumers can specify new product features

Design / fabrication

Consumer’s request are uniquely designed into the finished product

Design / fabrication

Each consumer order requires a unique design

Design / fabrication

Consumers can select features from listings

Assembly / use

 

Table 4.2: Overview consumer involvement factors Last.fm.

Type of modularity employed
The type of modularity that is employed is through standardization, in the form of adding options to standard products or interchangeability of components. The modularity factors that apply to Last.fm are that options can be added to a standard product, components are shared across products, and products are designed around a common core technology. This section gives examples of these factors. This section also explains which modularity type of Ulrich and Tsung (1991, in Pine, 1993) can be applied to Last.fm.

Options can be added to a standard product. When a consumer listens to music through Last.fm, tracks are automatically being added to the radio station, the standard product. It depends on the availability of tracks for the radio station how many different tracks will be added. The consumer is not able to manually add particular tracks to the radio station that he or she is listening to, but can influence the selection mechanism by giving feedback about the track that is being listened, or by specifying tags or artists to be added to the customized radio station. Feedback can be given in the form of just listening the complete track, skipping a track, tagging a track, recommending a track to another listener, banning or loving a track. Banned tracks would never be added to any radio station that the consumer listens to, while loved tracks are more likely to be added again.

Components are shared across products. Every track has the possibility to be found on radio stations that are musically different from each other, because listeners can classify each track with tags. When tagging, they do not choose from categories, but they tag the tracks as they like.

“In tag radio stations sometimes people mix very interesting things. Because of our similar artists they are more based on, like as I said, on peoples listening history. Sometimes as well it could sound different, but it has a similar feeling, maybe it has a similar mood to it.”

Products are designed around a common core technology. The common core technology of Last.fm is the Internet, the installed radio player and the available plug-ins for some music players. While the Internet is a common network for all consumers, the radio player and the plug-ins come in versions for Microsoft Windows, Linux and Apple MAC OS X. On this technology or platform, it is possible to add every available track, the digital product, to the platform.

All the above factors point to bus modularity, see Figure 2.2. It is the ability to add a module to existing series, when one or more modules are added to an existing base (Ulrich and Tsung, 1991, in Pine, 1993). It can be compared to track lightning or the Universal Serial Bus (USB). Table 4.3 gives an overview of all the modularity factors of Duray et al. (2000) that are the case at Last.fm.

 

Modularity factor

Stage in production cycle

Options can be added to a standard product

Assembly / use

Components are shared across products

Assembly / use

Products are designed around a common core technology

Assembly / use

 

Table 4.3: Overview modularity factors Last.fm.

 

Last.fm can be classified as an involver in terms of the mass customization classification of Duray et al. (2000). Consumers are involved in the design and fabrication stages of the production cycle, and the modularity type they employ is during the assembly and use stages. Listeners, or consumers, make use of unique products that consists of components that are shared around other unique products.

Possible variety

The possible external variety that Last.fm offers can be considered as very high. The possible variety is positively influenced by the multiple usages of modules, an interface that is not perceived as complex for the consumers and a platform that is efficient. At Last.fm, all three factors that influence the possible variety, influences the possible variety in a positive manner. Table 4.4 summarizes these influences. The minus sign at the second column indicates that the interface is not perceived as complex based on the constructed instrument.

 

Possible variety factor

Influences Last.fm

Multiple use

+

Interface complexity

Platform efficiency

+

 

Table 4.4: Overview of possible variety influences at Last.fm.

The variety of Last.fm comes in two different ways. The first is the total number of tracks which have been collected by the AudioScrobbler plug-in. This adds up to more than 40 million different tracks. These tracks consist of metadata about the artist, the track and the album. The second number is the total number of tracks that can actually be listened to. This number is a lot lower, but still almost one million different tracks that can be streamed on the radio stations.

To measure variety according to the metrics used in this research, some assumptions should be met first (Blecker et al., 2006). The first assumption is that the product family is built around platforms and modules. This assumption is met by the identified modularity factors. The second assumption is that the interaction process is carried out over the Internet, so that the consumer is involved in the production process. This assumption is also met. The third assumption is that the mass customizer has not implemented a cost calculation system that enables one to accurately evaluate the costs triggered by variety induced complexity. The implementation of a cost calculation system could not be identified, resulting in meeting all assumptions.

At Last.fm, the tracks are the modules, or the used components of the digital product. All components can be part of any digital product, which are the customized and personalized radio stations, because components are shared across products. The quote in Framework 4.1 shows the enormous potential of variety in radio stations that can be listened to on Last.fm.

“You can listen to your personal radio that is everything that you previously listened to. […] Then there is neighbourhood radio, which plays you things from your neighbours that you haven’t heard yet, which is like some sort of recommendation radio. Then, we have the direct recommendation radio, which takes your top artists and your neighbours’ and tries to find the best recommendations from both of them, so it takes a similar artist of your top artists and takes some artists from your neighbours and takes them together and this is the recommendation radio. […] Then there are similar artist radio stations, so where you can go to an artist page and listen to songs which are similar to Brian Eno, to say. And the way we calculate this similar artists is also based on our profiling data. Basically similar artists are calculated from people that listen to this artist also listen to that artist. That’s the way we come up with our similar artists, again from this data that we are getting from the AudioScrobbler plug-ins. […] And then we have artist fan radio, which plays you songs out of the profiles of fans of this artist, and then there is group radio, which plays you songs from the profiles of all group members, and there is tag radio, which plays you all the songs tagged with a certain tag by our users, and there’s a few other ones that I can’t even remember anymore. The most popular ones out of these are neighbourhood radio, and similar artist radio. […] And there are custom radio stations as well. If you click on the radio tab in the middle you can type in like three similar artists and it builds you a radio station based on these three artists.”

All possible variants of radio stations, or digital products, are a modular system where all tracks, or modules, can be present on every radio station. This indicates that the multiple use metric is very high, which enables high levels of possible external variety. The system also makes use of the submitted metadata from the AudioScrobbler plug-ins to add tracks to the radio stations, which positively influences the multiple use metric.

The interface of the Last.fm player is not too complex. When the player is installed, and the listener starts using it, the listener has to type in a similar artist or tag to start listening to the radio. All the options present on the installed player are straightforward. The radio station which is being listened to is being displayed, as well as the current artist, track and album. The artist, track and album are clickable and take the listener to the Last.fm website to the artist, track or album page, respectively. Other possibilities for the listener to interact with the player is to skip a song that is being listened to, to ban a song to never listen to it again, and to love a song to increase the chance of hearing that song or similar songs again. The listener can also tag a song on the player, or click on a link to go to the Last.fm website to write a journal posting about the song or artist.

While the features on the player are easy to use and are used frequently, there are some other possibilities on the website which are not used very frequently.

“If you look at your recommended music there is a slider. […] If you click on the top right where is your picture there is the recommended. Click there. So you get to the recommendation overview, and if you click on recommended music, you get on the right a list of recommendations that you can filter by popularity by dragging the slider at the top. […] This is a very good feature that nobody finds. So at the moment by default it is on weakly recommendations, which you can switch over to show overall recommendations then you get recommendations based on everything that you previously have listened to. You can fade away from the more popular stuff and drill down to the more obscure stuff. So if people want to find out about the more obscure stuff they can use this tool to get recommendations basically which is going to the more obscure section. Or they can turn it up the left hand side. […] A lot of people don’t know about this. […] It is one of the best things and we’re hiding it. […] It’s been around for a while but nobody finds it because it is very hidden.”

The last factor that influences the possible variety is platform efficiency. The product platform is made up of the radio player, which plays songs in a sequence defined by the interaction between the listener and the platform. This design suits every possible radio station variant, because every radio station just plays songs in a particular sequence. Changing consumer tastes and preferences do not have influence on this platform concept, which is making the platform very efficient. The only problem with this design is the lack of available songs that can occur for a particular radio station, which results in the player to stop playing.

The possible variety is positively influenced by all three factors. Modules can be used on every product, the interface is not too complex, and the platform is efficient.

Complexity and search costs

The analysis of complexity and search costs is divided in three parts. First, the perceived complexity will be analyzed, followed by the search costs and how perceived complexity is being reduced in the case of Last.fm in terms of customization.

Perceived complexity
The perceived complexity can be considered as high and low at the same time. The time that a listener takes to completely configure a product variant can be high, while this does not have to be a problem in the case of Last.fm. The abortion rate at Last.fm is also high and low at the same time, because there are reasons that consumers abort from using the radio player. Table 4.5 gives an overview of the perceived complexity factors at Last.fm.

For a listener to start getting musical neighbours, it takes about 100 songs in their musical profile. From that point, personalization can start to happen, musical neighbours are getting calculated and recommendations are given. During that first process, interaction with the player is happening, just by listening complete songs without giving any other feedback, or by giving feedback in terms of skipping a song for example or tagging a song or artist. It is not known if the listener is then satisfied, but basically they can influence the stations they are listening to all the time. After listening to about 100 songs, the radio player is getting better in terms of personalizing the station and giving recommendations. The more that is known of the listener, the better it gets in the long run, but the interaction process is a continuous and fluid process.

 

Perceived complexity factor

Influences Last.fm

Average interaction length of time

+/-

Abortion rate

+/-

 

Table 4.5: Overview of perceived complexity factor influences at Last.fm.

There are reasons for listeners to leave the interaction process or even to prevent from starting to use the player. The first entry barrier is that it is needed to download and install the radio player, which can be complicated for some people. The same is the case for the AudioScrobbler plug-ins. A reason to leave the interaction process is because personalization does not start right away. It can depend on whether listeners like the first few songs or not to decide carrying on listening. It is needed to invest some time to take advantage of the personalized radio stations.

Search costs
The search costs for consumers at Last.fm can be considered as acceptable. Since the data about consumers is not extensive and empirical data from consumers or users of Last.fm is not available, it is not easy to measure search costs very thoroughly. Last.fm is a social recommending system, where the listeners decide how to classify songs and artists by tagging them, and where discovering a product is more important than finding the right product. The number of alternatives searched can be considered as high, which negatively influences search costs for consumers. The product accuracy is good, and search agents are available but are not the primary way to find desired products. An overview is given in Table 4.6.

 

Search costs factor

Influences Last.fm

Number of alternatives searched

Product accuracy

+

Search agents

+/-

 

Table 4.6: Overview of search costs factors influences at Last.fm.

The number of alternatives searched at Last.fm is not entirely known. The product is the radio station, and the music played changes all the time. Therefore, the preferred product is not fixed, it is experienced while consumed. The radio stations can be influenced all the time, and other radio stations can be started at any time as well. It is not the purpose of Last.fm to reduce the time that consumers are interacting with the platform, quite the contrary. It is the purpose to let the consumer interact as much as possible and to discover new tracks and artists.

“Every tag is different, I don’t know how many tags, last time we looked there were 70,000 different ones but that was a long time ago so I don’t really know the current number but it is really interesting to scan through the tags because it’s really classified like… people listening to music is basically based on the music listeners rather than on editors or like some magazine journalists that are… which is very good, you can really see how music is listened to rather than how it is perceived from a commercial point of view when you’re talking about genres, because genres are essentially invented to present things better in the record store. You have to have something like this otherwise you can’t find anything. […] And it’s maybe much closer on how music is actually listened to rather than you know genrefication that comes from a point of view how music is sold or how music is best sold. […] There are endless discussions on how to write hiphop the right way, they all have their reasons why to write hiphop like this and to write hiphop like that, I tell you a 100 percent. […] If you combine this, there will be uproar in the community. They all have their reasons for writing hiphop like this or hiphop like that. Most of the people will say there is a specific reason why they call it something.”

The accuracy of the product can be considered as good. The easiest way to start using the radio player is by typing in a tag or an artist. When typing in an artist or tags, similar music is being played based on tags that are given to songs. Framework 4.2 explains how the accuracy of the product is kept high in terms of the tags that are used. Another way to keep the accuracy of the product high is by providing good recommendations, which is addressed later in this section.

The Last.fm website enables consumers to search for artists, albums, tracks, tags and labels on the music exploration page (Source: EM). This search mechanism searches in all submitted data from the AudioScrobbler plug-ins and the radio player. However, the Last.fm platform uses different strategies for consumers to enable them to find music or artists by their social recommendation system. It is not about searching, but about discovering new music.

Reduce perceived complexity
Last.fm offers various strategies to reduce the perceived complexity for consumers. The selection mechanism makes use of selecting by attribute rather than by alternative, default collections of music are offered, and consumer expertise is increased. Automatic as well as direct recommendations are made, and trust building and collaborative co-design is supported, see Table 4.7.

 

Reduce perceived complexity factor

Influences Last.fm

Attribute vs. alternative

+

Default version

+

Consumer expertise

+

Automatic recommendation

+

Direct recommendation

+

Trust building

+

Collaborative co-design

+

 

Table 4.7: Reduce perceived complexity, or customization metrics.

Last.fm presents their products by attribute, not by alternative. In reality, Last.fm presents the choice sets first in the form of popular tags, which then can be refined with related tags. This way of presenting choices also stimulates to learn preferences, by discovering music related to the choice that was made at first.

Last.fm makes use of defaults, for example by the recommendation system which can be customized by the consumer. By default the recommendations are set on weekly recommendations. This setting can be changed by the consumer to show overall recommendations, which will allow the consumer to get recommendations based on everything that previously was being listened to. This setting can also be customized by moving a slider on the Last.fm website between popular music and more obscure music. If consumers would like to find out about the more obscure music, they can use this tool to get recommendations which are more from the obscure section.

Last.fm offers possibilities to increase consumer expertise in various ways, and all these possibilities are available on the website. First, there are groups. Joining a group is a way for users who have a common interest to get together. Music statistics are generated for groups just like individual users, so listeners can see what musical taste certain groups of people have. Anyone can create a new group and encourage people to join the group (Source: FAQ). In these groups, people can communicate with each other about an artist or subject that the group is about. Listeners can learn from other listeners and increase their expertise this way. Second there are forums, where certain topics can be discussed, for example how to use the player. Third, there is the wiki on every artist and label page to write collaboratively about the artist. The purpose of the wiki is to provide a brief, concise, and unbiased description about the artist or label. This can include basic information about the person or group, the type of music they play, what they are most known for, and such (Source: FAQ). Fourth, there is a frequently asked question area on the website to increase consumer expertise.

Recommendation methods on Last.fm are presented both automatic and direct. Automatic recommendations are generated once a week. These recommendations are presented on the website, first on the personal page of the listener as a top ten of recommended artists. Methods for generating automatic recommendations are history-based filtering, rule-based filtering and collaborative filtering. History-based filtering, because the listening history of the consumer is being analyzed. Rule-based filtering, because the listener can apply settings such as weekly recommendations. Collaborative filtering because musical neighbours are included in the recommendation method. The user can listen to the personal recommendation radio immediately from the website. In the recommended music section the listener can customize how the recommendations are handled by a number of ways. First, the user can customize the recommendation radio by sliding a slider from obscure to popular songs, second by only recommendations from the last week or overall and third by dismissing recommendations.

“So in this recommendation section we also display things that are not necessarily on our radio, but you still find out about the artist, you still get artist recommendation. […] So the recommendations are also based on the metadata, instead of only the streams.”

Direct recommendations come from friends or groups. Listeners can recommend a particular track, album or artist directly to another listener, and to groups of listeners. Recommendations from a particular group that a listener is part of, can be enabled or disabled.

Trust or reducing risk is achieved by making use of a community. For the music this is achieved by tagging tracks or artists, automatic recommendations, and groups. It is the wisdom of the crowd that is important. Trust in recommendations is usually higher when the recommendations stem from peers (Piller et al, 2005).

“Our idea is that obviously the things that are most often used rise to the top of our charts. Usually the tags are quite good for a particular artist because most people use sensible tags. I mean with this knowledge of the crowd, the more people use a specific tag then obviously it has more relevance, so we usually weigh things, so it might be that one tag that one person has used for a particular artist is maybe not likely to be prominently displayed, but that’s the thing. The more people use a certain tag, the more prominently it gets displayed.”

Collaborative co-design is done by a continuous interaction between the listener and the Last.fm platform. The listener can start building a personal profile by listening to some predefined stations, such as global tag radio, or by submitting tracks by using the AudioScrobbler plug-ins. After about 100 submitted tracks, the Last.fm platform has enough information to start generating musical neighbours which enables the listener to discover new music that is probably something that the listener likes.

For consumers, the perceived complexity can be high and low at the same time. At Last.fm, much is possible for consumers to reduce the perceived complexity. Instead of decreasing search costs for consumers, Last.fm stimulates to use the platform as much as possible to discover new music.

Lessons learned

Last.fm can be classified as an involver in terms of the mass customization classification of Duray et al. (2000). The earliest point of consumer involvement is in the design and fabrication stages, while the type of modularity employed is present in the assembly and use stages. Variety at Last.fm is high for consumers. All three factors which positively influence perceived variety are supported, while at the same time all assumptions (Blecker et al., 2006) to measure variety according to these factors are met. The perceived complexity at Last.fm can be considered as high and low at the same time, the search costs for consumers at Last.fm are acceptable and Last.fm offers various strategies to reduce the perceived complexity for consumers.

Last.fm tries to let the consumers interact as much as possible with the interaction system and the product accuracy is good. This also has implications for the number of alternatives searched by consumers. According to theory, a high value of this metric has a negative influence on search costs. At Last.fm the products change all the time, which makes it necessary to keep searching for alternatives. Search agents are not the primary way to search for products. The whole system is about discovering new products by listening to recommended music. The social driven community of Last.fm plays an important role here. All factors that lower the perceived complexity could be identified, which can be seen as ways to support customization.

Advertisements

One Response to Case study Last.fm

  1. I know this if off topic but I’m looking into starting my own weblog and was curious
    what all is needed to get set up? I’m assuming having a blog
    like yours would cost a pretty penny? I’m not very internet savvy so I’m not 100% certain.
    Any tips or advice would be greatly appreciated. Many thanks

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: