Archive | October, 2012

The iLike AdSongs Platform

19 Oct

By the end of summer 2007 had mostly regained ops stability and a workable hardware installation schedule following the growth of Facebook’s platform launch on May 25 – 4 million new users over the first two weeks alone kept us adding hardware and fixing software continuously for a while, changing the wheels and engine every few days on our car while it raced. Our original standalone social network, continued to grow, but our growth on Facebook was vastly outpacing our organice user growth, and in the subsequent months our syndication to other social networks like Hi5, Bebo, Orkut, and MySpace would double and double again our user base and our reach across the web.

On all the social networks and websites where we projected our “widgets” and “app” we used a common visual element, themed to its surroundings, which allowed in-line playback of full songs or 30-second samples (depending on rights). We called it the “Song Expando” internally and the amazing Steve Rider crafted it with tremendous love. It was a tight little piece of javascript & behind-the-scenes Flash that showed animated buffering and playback progress, auto-played to the next song if there were several in a group, and could open up slightly in-page to provide artist and album information, links to videos on YouTube, purchase links through iTunes, and user- and geographically-targeted concert ticket buying and social/sharing information.

In November ’07 after plotting the growth in the number of Expandos we were rendering, how many tracks we were playing and how many MP3 and ticket purchase referrals we were generating, I wrote an internal memo proposing using the Expando more explicitly for revenue by syndicating its use around the web and selling the ad space, making it a Google AdWords-like product available to third parties, generating revenue and sharing it with sites (publishers) and the content owners (artists and labels). I ran some tests where I placed keyword-targeted AdSense ads within Expandos to see what they could support – they achieved CPMs and CTRs so high that our Google contacts thought we were doing illegitimate click-fraud until I explained the circumstances and they examined the placements. Google really liked these units and gave us an exception for their format since they only appeared after the user pushed play and in some cases were within an iframe – normally both forbidden by AdSense. For me, the performance of this ad format crystallized my view on the importance of user attention and the physical placement and interaction with ads in a workflow (the Ad Contrarian said this recently better than I have yet). More on this subject, someday.

The technical work to syndicate Expandos at web-scale given our architecture and existing social-network-scale was trivial. The rights environment and complications around revenue sharing were (and are) hideously more complicated. You see, we were operating in a grey area with the main rights owners. The major US record labels – but especially Universal Music and Sony-BMG which held about 60% of the market – didn’t much like our use of music clips. (Actually, they liked it just fine; they didn’t like the fact that we didn’t pay them directly for their use – we paid Muze and others for clips). Their artists all loved iLike and loved having their tracks available (mostly as 30-second samples) on Facebook and around social networks, because their own contracts placed onerous restrictions on how they could distribute samples and tracks online. iLike was a watershed for many artists and small labels in terms of getting new fans through music discovery and being able to reach out and communicate to fans. At the time, iLike had more active artists and a larger “followers” graph of fans than Twitter by several orders of magnitude, and was on-track to surpass the then-dominant fan graphs on MySpace. From the moment of our launch in October 2006 we had been playing “Clip Chess”1 over the use of 30-second samples on and our syndication to social networks, especially Facebook, triggered a huge escalation in the game. Adding revenue generated more directly by the clips into what had become detente would have been a dangerous gambit and an invitation for lawsuits.2

So we held off simply launching AdSongs and instead continued negotiating with labels while opening conversations with Apple,3 Rhapsody,4 Microsoft, and Amazon about whether we could safely use and syndicate their music store samples and get out from under the shadow of the labels. After an extremely painful integration with Rhapsody over the spring and summer of 2008 we shipped a neutered form of AdSongs (lacking revenue sharing and without ads in Expandos on 3rd-party sites) in September ’08. It took the form of both a self-service ad network for concert promoters featuring pay-per-interaction ad-units across our properties and a developer platform for embedding songs and playlists (and playlist builders) into other Facebook apps and onto websites. These tools became fairly popular with developers and promoters, but the horrible technical performance of Rhapsody and the continuing take-down requests from labels (despite RealNetworks’ and Rhapsody’s promised safety umbrella) kept us from investing in AdSongs as a revenue source.

Rhapsody’s cripplingly bad streaming performance and outages, horrible user conversion to subscriptions and the continued user and partner (especially Facebook) feedback that we needed more and better clips internationally as well as full streaming led us to finally take the necessary first step with labels: MP3 sales. Our belief was that 80% of the technical effort – payment processing, fraud-detection, content-ingestion, hosting, and delivery, promotional requirements, per-label volume reporting – would be completed by taking this first, legal-boilerplate 10% effort step of setting up an MP3 store, and that good MP3 sales volumes would help our much more complicated 90% legal and business negotiation effort for US streaming rights. We were correct about the technical effort – it took us only about 6 weeks across a few people and me at about 75%-time5 – but the legal and business development effort was not entirely boilerplate and took a great deal more mental time and effort for Hadi to tee up – almost 9mo of effort and several trips to NYC and LA, shipping in late summer of 2009 while we were already deep in MySpace acquisition talks, where we had hoped to have it live by February.

So in the end, after negotiating and building an MP3 store could we finally have built and shipped an AdSongs product syndicated around the web? Sadly, not really. For starters, a great deal of our extra negotiation time was in fact tied up in the fact that iLike was already syndicated to a set of properties that were not “within our own domain of,” and this triggered a visceral negative reaction at the labels who believed that they were due a share of the overall  revenue of any site that supports playback of their music, not just ads associated with the playback of their music, not just ads elsewhere on the pages containing music, either. You can actually understand a great deal about negotiations with record labels and the history of music on the web once you internalize what I just wrote: they believe they are due a share of the overall revenue of any site that plays back their music anywhere.  Our rights were were in the end structured to allow putting clips and buy buttons only within the specific properties we already had and other uses were up for individual review. What their philosophy meant for the concept of AdSongs – that they would share in the ad revenue from the contextual ads served during playback of content as well as the ticket, MP3 purchase, and other sales directly in the Expando, and that so would the site owner and iLike – was that it simply wasn’t something they were open to or ready for. While negotiating the MP3 store, we simply didn’t feel we could bring shared advertising revenue into the business discussion unless we wanted to spend years negotiating.

So, after much ado, here is the AdSongs proposal from November 8, 2007:

AdSongs: Because Content Wants to Be Free but Eyeballs Never Will Be

Billions of web-pages across millions of web-sites large & small contain links to music content which can be played on-line or downloaded by users. Some of these sites & pages (including all iLike content) link to legitimately licensed 30-second samples and full-song clips for free streaming (the majority) or free download (the minority). The problem is that the legitimately licensed content is not pervasive enough; it is difficult for both large sites and small consumers (music bloggers, for instance) to acquire rights to use music content legitimately. A further problem is that the one-time, high-cost, up-front licensing model for samples and full-streams employed to date sets up a frustrating long-term revenue imbalance for the music content owners – the lack of per-listen revenue for technical and other reasons from licensees makes music content owners desire on-line licensing even less. So, not only are many useful promotional opportunities lost (blogging, news articles, email) but the barrier of licensing in time, money, and infrastructure requirements is causing more small-scale and large-scale use of illegitimate content. This unfortunately feeds back on itself to further dissuade legitimate licensing and dissemination of music content. And this is a shame. Because, at no time in history has music been more in-demand. As we have seen at with our exponential growth on social networks like and Facebook users crave music on web pages. Users want to interact with music. They want to recommend music to one another. They want to read about music, about musicians, about upcoming concerts, about album releases. They want to review music, write about it, comment on individual songs, and create playlists. They want to email music to their friends. And while they are reading, emailing, and creating they love to have a non-interrupting means of listening to the music, sampling new music, and finding out more about that music.

AdSongs: Musical Hyperlinks + Contextual Advertising

What if you could make legitimately licensed music available everywhere on the web in a format where it is easier and more natural to use than illegitimate music? What if users sought out this format of music and preferred it because it gave them the best streaming & sound experience right in their web-pages without interruption, allowed them to easily buy the song or album if they liked it, and allowed them to share their music discoveries with friends? What if web-page authors, from the smallest stay-at-home-dad-blogger to the largest commercial web-site could add this form of music to their pages without having to negotiate a complex license? What if web-page authors were furthermore incented to add this format of the music instead of illegitimate links because it made them money? Finally, what if every time a user on any web page listened to this music the artists and content owners also made money? It’s not impossible. This is just what we’re building at iLike with AdSongs. Think of them as the musical equivalent of the hyperlinks you click on every day as you navigate around the web. You can see the initial form of our AdSongs “musical hyperlinks” on, Facebook, and in search results (Figures 1 & 2). These existing iLike AdSongs appear in over 300 million pageviews each month by 20 millions unique users. AdSongs already present in-line local concert indicators (a form of contextual advertising) to specific viewers based on their geographic location. In just 6 months AdSongs have grown to become the #1 referrer of music purchases to Apple’s iTunes Music Store and the #3 referrer of ticket purchasing to Ticketmaster. But this is just the beginning for AdSongs.

Figure 1a. The popular songs module on U2’s Artist page on ( before the user has chosen to listen to a song.

Figure 1b. The popular song module after the user has chosen to listen to the song Vertigo – the AdSong expands to present purchase information.

Figure 2a. An AdSong (currently “static”) is presented in the right-rail when searching for musicians on (

Figure 2b. Static AdSongs like these from Billboard syndicated charts ( are shown in over 20M pages on Facebook. iLike will be transitioning these to dynamic “expando” AdSongs at the beginning of 2008.

AdSongs offer a unique opportunity for highly contextualized advertising because they are an in-visual-field control surface that draws a user’s eye and invites interaction. Unlike banner and edge-ads, dynamic AdSong expandos appear when and where the user interacts with them, whether they are in a song-list (Figure 3), in-line with text (Figure 4), or in a playlist, chart, or album-view of a song collection. AdSongs expando ads are initially targeted based on the song being played and the geographic location of the user playing them. In Q1’2008 we will add targeting by standard user demographics if the viewer is one of iLike’s 15M members across and Facebook as well as targeting based on their specific musical preferences which are unique to iLike’s knowledge of the Artist Fan Graph.

Figure 3. A blog review of a concert containing a grouped playlist of songs from each of the sets played by the concert lineup. Only when the user clicks to play a song and listens to the content do ads show.

Figure 4. A an article reviewing Led Zeppelin II contains multiple AdSongs “music hyperlinks” in-line with the article text. AdSongs expand and play without intruding on the reading experience of the user.

For small-scale bloggers and other web-authors adding just a few songs, AdSongs will be as easy as searching for a song on and copying universal HTML content into their pages. For commercial bloggers and larger commercial web-sites, they can sign up for an AdSongs account with iLike and receive an AdSongs-Key so that they share in the revenue from AdSongs appearing on their pages. The very largest-scale sites & music stores may choose the additional step of paying for the placement of their own songs or to have no advertising at all in their AdSongs. In all cases iLike’s AdSong engine takes care of dynamically choosing and delivering ads when songs play, streaming those songs, tracking impressions and click-throughs and distributing the revenue to content owners and publishers.

Revenue Sharing

Web publishers (web-site owners) are used to seeing anywhere from a 50:50 to an 80:20 (their advantage) split of the advertising revenue from ad networks when they relinquish page space for ads. AdSongs are a different types of intentionally placed and integrated content – when the user pushes play, they open and play music content that is owned by a label or artist, content that should be paid for and which contributes significantly to the efficacy of the advertising within the opened space. Therefore in a typical case – specific songs that a blogger places in an article, a selection of specific songs in a custom playlist, for example – we  propose that 40% of the advertising revenue goes to the song content owner, 40% goes to the publisher, and 20% goes to iLike to support and operate the AdSongs infrastructure. The combination of (a) the higher click-through rate on ads within AdSongs which are opened, (b) their ability to highly target advertising to individuals based on the songs they interact with and thereby command higher CPMs, and (c) the fact that they offer in-line display and don’t consumer a publisher’s page space, suggests that AdSongs revenue will be additive for publishers. So we believe this revenue-split is justifiable. We do anticipate that there will be many other AdSong modules for publishers representing more space-consuming elements such as playlists, “top-10” charts, artist-specific ads with playable content, and even label- or artist-promoted results in search-engines which may have different revenue-splits, more advantageous to publishers because they are part of label- or artist-promotional efforts.

1 Clip Chess is complicated and interesting enough that I’m pulling it out into its own post soon.

2 In hindsight I think it would have been better to push the envelope with labels sooner so we could make a go/no-go about clips and full-songs and about the overall business of iLike rather than letting the issue bleed us slowly like a leach. Fail Fast, Fail Cheap, Fail Often is a better plan than Succeed Almost, Limp Like A Zombie.

3 Our almost-acquisition by Apple in February 2008 is a horribly painful story for another day.

4 Or as we liked to call it, Crapsody. Also a story for another day.

5 Granted, among the people working on the store were Eugene Zarakhovsky and Seth Cousins, two of the most gifted and unbelievably productive engineers I have ever had the pleasure of working with.