Yesterday morning I realized something that had been floating around in my mind for a while: Niche choices provide a whole lot more insight than mass ones.
I was playing with last.fm’s similar artist radio. The basic premise is you put in an artist name and it plays you comparable music based on the data they’ve collected from users. Thing is, the similar artist player spits back terrible results if you put in an artist like Coldplay. That’s because for a band that popular, similarities in people’s other tastes don’t necessarily mean similar sound. The pool is too diluted.
That makes perfect sense logically, but for whatever reason yesterday morning it struck me as incredibly important. When I put in the Guther, the system returned some very accurate similarities as well as some interesting, but good, deviations. That’s because for an artist like Guther, who very few people listen to, the other artists people listen to are more meaningful.
None of this is to take anything away from popular bands like Coldplay, I actually like them quite a bit. Instead, it’s just to make the point that when a person makes a conscious decision to consume something niche it says much more about their taste than a mass artist/movie/etc.
Let’s try it another way: Imagine going into a room and asking everyone who’s visited Yahoo! to raise their hand. Everyone in the room would have their arm in the air I assume. Now ask who’s visited NoahBrier.com. Pretend one other person raises their hand. (Come on . . . use your imagination here!) The odds that you have something to talk about with that one other person in the room is far higher than you having something to talk about with everyone in the room.
I think this explains two important pieces of this 2.0volution, specifically as it moves outwards towards the general, more diverse, public:
- The best recommendations come from niche choices.
- In order to compare niche choices effectively you must have scale.
What’s scary about that statement, is that it seems like as soon as these things go mass they will no longer be useful. That may be true for something like ‘similar artist radio’ for Coldplay. However, what a system like last.fm can do to get over that issue is look at your entire library and compare from there. Thus their other feature, ‘recommendation radio’, is even better. By looking at everything you listen to and comparing it to everything everyone else listens to, you get some pretty insightful picks.
None of this is new, Amazon’s been doing it for years with their ‘people who bought this bought that,’ but it’s still a really big deal.