User Generated Content
We spoke briefly in class about website development and utilizing technologies such as AJAX to enhance user experience. Our class has a mix of students who have experience working with some of the popular web frameworks.
If you have experiences coding in web frameworks – or if you experience working with or managing teams of coders who have worked with different languages and frameworks, please share insights and lessons learned in the comments.
We are fortunate to have Paul Vidich, WG ’81, graciously contribute to our blog regarding his thoughts on music and startups.
Paul is currently advisor to numerous startups (Brightcove, Reverbnation) and Betaworks, an innovative startup and technology development firm, and formerly was head of digital strategy at Warner Music. For more on Paul, check out his LinkedIn and Crunchbase.
“Off Notes: Online Music’s Bad Record of Investor Returns” by Paul Vidich, betaworks advisor
Comments on Investment Opportunities in Online Music
The online music space continues to attract enormous numbers of users, but the underlying technologies are relatively straight forward to implement (audio streaming, database management, etc), which has meant the field of competitors continues to be crowded. In looking at the list of companies now seeking investment, I’ve aligned their activity against some business models in order to access their attractiveness. This analysis seeks to bring some historical perspective to the continuing failure of the online music space to generate meaningful investor returns.
First, two principles. It’s important to understand that, for consumers, ‘it’s all about the music.’ Consumers want to listen to, interact with, or own music. Derivative activities, such as information about tours, networking among fans, etc., are secondary drivers of interest. Second, music rights, which are available broadly under statutory licenses or exclusive licenses, come with costs and restrictions that limit a music site’s program flexibility and gross margin potential, which means that, except for Apple, no online music site launched in the last ten years has been profitable. Pandora, one of the most popular with 2 million daily users, and launched in 2000, continues to lose money. Whether it makes money or not will depend on the outcome of Congressional intervention into the copyright royalty tribunal rate that was set last spring.
Music Focused Sites:
Online radio sites, like Pandora, are collectively the most popular music destinations on the web. This category includes Last.fm, Slacker.com, Jango, anywhere.fm and many others (there are thousands). Rules that control how music is played are defined under the US statutory license, and they severely limit interactivity: a play list must consist of at least 25 songs from six artists and no more than two songs from any given album. It’s radio like. It’s ad supported, but the ad model has been somewhat problematic. Users at their PC minimize the browser while listening, and they do other things while a song is playing. Each time a consumer clicks on a new radio/genre station, the player opens with a new ad presented. A typical user might listen for an hour and in that period click 3 or 4 times. This puts an upper limit on the revenue potential.
Some stations, e.g. Pandora and Slacker, allow a consumer to upgrade to a premium station that offers unlimited song skips and no ads, (for about $4 per month), but it hasn’t been terribly popular. Some of these sites have added social networks or tou r information, but those features have not added significant value to users whose principal interest is music listening.
I would characterize this space as capped revenue upside, relatively high content costs, and limited potential for product differentiation among the large number of competitors.
Several sites focus on providing consumers with access to information about music and musicians. Songkick provides geo-targeted information on live music — someone in London or New York can find which bands are playing on a given night. Their database is structured to search by artists, city, date; and widgets have been created to enable music bloggers to embed the Songkick code with20any artist mentioned in the blog. This is a niche play, ad supported, which is more likely to become a product feature in one of the larger music community sites, or an online radio site. Large databases of music information do exist, principally, AMG (All Music Guide), Gracenote, Web Music Database and Ultimate Music Database. These services vary in cost, but operate both as b2b plays and consumer facing sites. I’m not sure if there’s room for a new service that could scrape music data from music web sites and to build a more efficient, robust repository of information. It might be worth looking into, so long as the underlying technology might be used for a variety of purposes: recommendations, other content types, etc.
Music Social Networks
Myspace’s rapid initial success hinged on the premise that it conn ected and artists to their fans, and fans reached across artists to connect with each other. Tens of thousands of bands formed communities among millions of fans who wanted to have a relationship with the musical artist. Myspace continues to be a significant player in music community, but other sites, specifically Facebook, have won over consumers by enforcing design/content order into what became a cluttered Myspace experience.
Hype Machine, a new entrant in the music social network space, builds music intensity awareness from the music focused blog sites, and offers a non on-demand radio listening experience that operates under the Copyright office compulsory license. Hype Machine’s idea of being the ‘smart’ place to discover new music is a pretty solid. The challenge for music social network sites is that there are several in the market, some targeting ethnic groups (such as Black Planet) and the unfiltered, often unmanaged editorial content, can be hard to sell ads against.
Subscription services that enable tethered portable on-demand listening to a music library have generally failed. AOL and Yahoo left the business. Napster was sold to Best Buy for 1/3 of gross revenue, and still loses money, and Rhapsody I believe continues to lose money. Nevertheless, consumers love Rhapsody. The problem is that the content tax from the music labels takes about 60% of the revenue (and there are high minima), with the result that these direct marketing companies aren’t left with sufficient gross margin to profitably acquire subscribers and pay overhead. Label content cost minima force a minimum consumer price of $12-$14 per month, which is ludicrous for a music service. Subscription services work only if the price is in proper relationship to a compelling consumer value proposition.
In summary, the online music space is one that investors should approach with caution. Some new sites have interesting product features, but not a sustainable business model of scale. It’s important not to forget that illegal p2p file sharing continues to be a major way by which many young people obtain music. The music labels Achilles heel is that they are increasingly becoming dependent on online sales, but they have not yet figures out how to effectively, with scale, market and promote online. This creates one investment opportunity. An online site that can create viral tools for the implanting and dispersal of music information, or link to music, has a shot at recasting the leverage in the industry.
Data mining techniques – Bonferroni Rule by Rob Sebastian (LinkedIn)
Gautam suggested yesterday in class that, in data mining with a large number of potential variables, there is a risk that 5 out of 100 will be randomly assigned statistical significance based on sheer chance under a 95% confidence interval (p < 0.05). Kartik correctly suggested that you can / should use your intuition to weed out potential overfitting, but I also wanted to follow up on my comment from class. There is a method – the Bonferroni Rule – intended to correct for such over fitting when working with a large number of potential variables. According to the Bonferroni Rule, one can reject H0 if p <•/n, where a is the desired level for the test of H0. So in the case of our 95% CI, a = 0.05 and if you’re considering 100 variables then p must be less than 0.0005 = 0.05 / 100 for inclusion. This more-stringent threshold for inclusion helps correct for the sheer number of variables you are considering when data mining. The linked .pdf is from an excellent lecture by Professor Bob Stine on Over-Fitting from his STAT 622: Statistical Modeling elective.
Comscore US Digital Year in Review
Pablo Lema (LinkedIn) found this great piece of research. Enjoy!
Facebook is reportedly preparing to launch a fully functional email system, complete with POP and IMAP support (so you could access your facebook email from outlook, your iPhone, or any other email client). TechCrunch has more details here.
What are your thoughts? What is Facebook’s strategy here, and how might this move impact profitability?
Do you plan on using Facebook email – why, or why not?
Leave your thoughts in the comments.