Sunday, March 19, 2006
Pandora and Last.fm: Nature vs. Nurture in Music Recommenders
This is a must-read analyis. (Hat tip : Paul Kedrosky)
A few quick points:
1.Really cool example of edge competency is how Last.fm's plug-in allows it to automatically capture data to improve its search filters. If you have followed Umair's thesis so far, you should be able to immediately see increasing returns to scale through positive feedback loop and how it creates a very strong barrier to entry. This is strategically brilliant.
2.This also illustrates a theme I blogged about earlier.
We seem to have two distinct strands of how edge competencies are developed : algorithmic or anthropocentric (yes, I am coining that term here and now). Although they both have common strategic roots, clearly the execution will be very different. Firms will need to pick one or the other and allocate their resources accordingly. An implication is that firms must clearly know what they are doing : they are unlikely to succeed with a hybrid approach. One way to see this is that making this choice offers a migration path for the firm's existing core competence. In other words, firms can see where they are today and then make the most doable choice. For example, it will be an idiotic strategy for Google to develop an anthroppocentric edge competency, because they are all about computation and algorithms. Witness how miserably Google's social network attempts have failed.
3.Gratutious strategic advice for MySpace : please shut up and buy Last.fm already!
Agreed. There was a really interesting post on Wired news the other day about Digg vs. Google News, pitting it as a John Henry-like match of man vs. machine.
What's interesting is that as we are able to harness the power of multiple humans at once, the wisdom of crowds come in play and lends the immense advantage of trust and relationships to the whole thing.
Ray Kurtzweil's newest book, The Singluarity is Near, predicts that very soon machine intelligence will surpass that of humanity and then vastly surpass it. Not that big a surprise. Look at what happened with Big Blue & Kasparov.
My question is will machines ever find a way to make us trust them? Will there ever
be a time where AT (artificial trust) becomes a reality?
Last.fm has done a great job of being first-to-market with the automatic feedback-loop concept. And its surely a strategic hit.
Ultimately the social network recommenders and machine-based recommenders will merge. This is inevitable, IMO.
Note other players are moving into the Last.fm space, with a similar strategy. PodOmatic has released a social network to complement their free podcast hosting services. (http://www.podOmatic.com)
They've got a client download enabling networkers to share their podcast and music favorites automatically. Similar to last.fm, but in many ways much better.
I think digg does a great job attracting the news and readers in the tech world.
It seems to me that business success on the web is about 1. social networking 2. useful contributions.
Here's an example of this: www.1000greatthings.com.
It's like digg but for the non-digg crowd. And it's early days yet!
Check out our entry into the LastFM / Pandora ring...
Tagworld.com music recommendation engine!