I wrote on Wed about the notion of “total attention” as a useful metric by which to evaluate music or other websites — but didn’t explain my thinking or actually run the numbers.

As so many have written over the last decade, a large and growing proportion of GDP comprises products and services that are (or can be) rendered in bits. As it becomes ever cheaper to create, copy and distribute bits, scarce resources in production (broadly defined) are increasingly giving way to scarce resources in consumption – i.e. consumer attention. As an aside, this is why I’ve long believed (and Chris Anderson eloquently wrote a few years back) that the most significant driver of value for digital products is really just “matchmaking” or filtering or discovery. That is, helping people find those bits of greatest interest from the digital haystack, so to speak.

In any event, if consumer attention is the most precious resource in the digital value chain, then its proper measurement is critical. Further, I’d argue that the more attention a web offering can command, the greater the opportunity for monetization, all else equal. If a person visits more pages on a website, listens to more minutes of a webcast, or uses another web-based service more frequently, there are more opportunities to place promotional or branding messages, whether priced on a CPM, CPC, CPA or other basis.

Likewise, a user’s choice to spend more of his “disposable” hours in the day on a website/cast/service also suggests that he ascribes greater value to it, and — again, all else equal — there is a greater likelihood he may be willing to pay directly for some aspect of that product, whether priced on an a la carte (commerce) or ongoing (subscription) basis.

To measure the total attention that a web offering commands, 3 factors must be considered:

  • unique visitors
  • frequency of visits (e.g. avg visits/unique)
  • average “stay” per visit

While average stay per visit (e.g. minutes on site) would seem a logical measure to assess the magnitude of a visitor’s activity on site — since it tracks the total minutes of attention granted — another, perhaps more practical measure is page views per visit. Page views per visit is useful b/c it (1) generally corresponds to new ad placement opportunities (with each page refresh), and (2) ensures that a user is “active” on the site and does not simply have the browser open in the background while, say, writing an email.

So then total attention can be defined in 2 ways:

  • unique visitors x visits/unique x minutes/visit = total minutes of attention
  • unique visitors x visits/unique x pages/visit = total pages of attention

I used compete.com to rank the top 10 music sites on the latter basis (the former really won’t make sense until online audio ads move beyond a nascent stage). As this post has gotten a bit overlong, I’ll include these rankings in a new post.


If one assumes advertising (or other sorts of attention-based business models) will be the primary driver of revenues for online music services — or really any online services — then a total measure of attention is useful in establishing the value of the service.

I’ve generally presented music sites ranked by “people” (compete.com’s version of unique visitors). But really a total measure of attention should reflect:

  • unique visitors
  • frequency of visits (e.g. avg visits/mo)
  • average “stay” per visit

I checked a few of the top music sites on the average stay measure and found some intriguing results:

I’m guessing that some of the reason Live365 ranks so highly has more to do with its user base (older, more passive) than specific site features. But it is somewhat surprising.

I haven’t updated the music 2.0 rankings for quite a while. There’s some very cool (albeit legally questionable) new players on the horizon (Muxtape and Mixwit), and several existing sites have demonstrated serious growth (like Qloud, just acquired by Buzznet, and Jango, which seems to have come out of nowhere).

This time round, I’ve dropped the two-point-oh designation and picked up old-schoolers Rhapsody and Napster, increasingly useful now that many services are offering legal on-demand access (e.g. Last.fm now does on-demand in addition to radio, imeem now is licensed to offer on-demand access). I wanted to include the music offerings from Yahoo, AOL and MSN but can’t seem to get subdomain data.

I’m probably still missing some sites that ought to be on here. I know that MySpace and Buzznet (and, to a lesser extent, the other big social networks, most of which have some sort of music component) could arguably be included too. I’ve kept imeem because music is such a big component to their service; MySpace is the tacit #1 service.

Lastly, I’m not sure what’s up with the retroactive change to imeem’s numbers but it’s now showing a huge increase in prior months over what had previously been reported.

I’ve been playing around with Tumblr. While I’m no doubt a bit late to the whole tumblelog concept, I’m definitely liking the casual-ness of it. No need to formal titles, ready ability to add different content types, and useful integration of feeds from other social apps (del.icio.us, twitter and, in fact, wordpress). It also looks nice. And is based in NYC.

Accordingly, I’m going to increasingly shift the focus of Sampled + Sorted to work (already have to large extent) and a new venture (called 8tracks) that will launch in limited beta next month. Meanwhile, I’m going to use my new tumblr page to aggregate my social app feeds, post music and photos, and do anything else that’s more on the personal tip.

WordPress for work, Tumblr for play.

A research report from AccuStream, published yesterday, found that total listening hours (“TLH”) for the internet radio sector in 2007 were 4,850,000,000 — a 26% increase over 2006. And I suspect this is only for the US so would exclude, for instance, a large part of the listening on Last.fm.

Internet radio revenues, however, were estimated at a mere $92,000,000 — not so impressive. A quick bit of math is revealing: $92m/4.85bn = $0.0190 per hour. That is to say, revenue/hour was just shy of 2 cents. Unfortunately, the rate for digital sound recording performance royalties (the thing I wrote about a lot last year) is $0.0011 per performance (per track, per listener). Based on an average of 14 tracks/hour, this equates to an hourly cost of $0.0154. In other words, the sound recording royalty by itself consumes 78% of advertising revenues.

Assuming these figures are accurate, the internet radio sector is paying the labels a 78% share of their revenues. By contrast, traditional (aka terrestrial) radio pays 0% of their revenues — nothing — to the labels. And satellite radio pays 6-8% of its revenues to the labels.

For internet radio services, the remaining 22% of their revenues must be divvied up between the musical composition royalty (probably 4-5%), bandwidth, employees (which is typically the most expensive component of cost), overhead, and, well, profit. Or not. The math doesn’t work.

Now, admittedly, there are a couple of problems with the above analysis. First, this study seems to cover the US only and the internet is, by definition, global, so both TLH and revenues are far too low. Second, oddly, AccuStream makes no mention of normal visual ad revenues (banners and the like, not video), which is a significant source of ad revenues.

One other point that’s worth noting: the average price (CPM) of ads placed in/around internet radio is clearly lower than it is for terrestrial radio, in the same way that the price for ads on the internet (in general) is still considerably lower than it is for ads placed in traditional media. This will change over time, but it’s too bad that many internet radio services won’t live to see this day, should current rates stand.

Just a quick note:  from the graphs below — see my post from Dec 14th on the top 20 music 2.0 sites, including graphs that automatically update thanks to compete.com — it looks like imeem has begun its upward trajectory again and, coupled with another decline at PP, is now in the #1 position.

Other big trajectories:  MOG is now approaching the size of Live365, Qloud has jumped significantly (now at nearly  225k uniques/month), and SoundFlavor (which features a PP or Seeqpod-like streaming MP3 search engine) has seen two solid months.

As many of you know, I’m a big fan of Last.fm — it’ll be 2 years ago tomorrow since I first posted about it, and if I weren’t focussed on my own venture, there’s no place I’d rather be working. So yesterday’s announcement is definitely interesting: users of Last.fm can stream any track up to 3 times gratis, and thereafter they can either (now) buy the download or (soon) subscribe to a Rhapsody/Napster style service to continue to stream on demand.

I think the positioning of the announcement was a bit unclear, however. This is NOT a free, on-demand service like imeem or lala; rather, this will ultimately be a subscription-based, on-demand service with a free “sampling” component very much like what Napster offered a couple of years ago. In fact, Napster’s offer was (is?) actually for 5 (rather than 3) free listens before you had to buy or subscribe, and I blogged the underlying math at the time. Quincy Smith said they wanted to offer more free listening, but it remains to be seen whether the Majors will be willing to grant a much higher threshold.

Fred Wilson also suggests the price for Last.fm’s subscription will be $3, but I’m pretty sure this is just the pricing on their current premium (ad-free, personalized) radio sub service, and I seem to recall having read that the on-demand subscription will be higher. I think it’s unlikely the labels would grant Last.fm a less expensive arrangement than what Real, Yahoo and Napster get, except as part of a comprehensive deal, including upfronts (which could be the case, we’ll see).

The reasons I see a Last.fm subscription potentially faring better than Rhapsody/Napster/YMU are (1) the on-demand functionality is layered on top of a free service that is already highly successful, with a large base of users (20m registered users), and (2) the service has a core social dimension which was lacking in previous efforts.

I like Felix, Martin, Spencer and the rest of their crew and applaud their efforts to make this model work.