Big catalog = infinite specialty shops

My perspective is that the problem with this IS seo.

The article talks about buying 1,000 domains. One big seo factor is domain authority/trust. With 1,000 new domains, you have to build up 1,000 separate properties with authority/trust, that's no small feat as anyone who's tried building up 100 domains at the same time will tell you. The common tactic here is going after the long tail, which is the only way I can see this working and still staying within a sane budget.

Beyond seo the article talks about marketing in general, and how great it will be to market laser targeted niche stores, all 1,000 of them.

So if I've spent a couple of hours finding a great Japanese music site that is PERFECT for listing my Japanese niche music store/domain, then I've just spent a couple of hours on 1 domain, only 999 more domains to start marketing! Yay!

That's horrific.

If you take the approach of a niche property within the same domain, all of your seo and marketing efforts for 1 niche store will transfer over to the rest of your properties, since, of course, you'll inner link them all from within the same domain. This also works great from a long tail perspective.

The marketing/seo execution of 1,000 separate domains is the challenge...

I find it odd that there is so much enthusiasm for this idea in the comments on Sivers' blog because to me it seems a bit outdated.

And I agree w/ coffee and sabat that sure, anyone can purchase and setup a bunch of niche sites w/ individual domain names, but gaining trust (both from SEs and from consumers) is the key factor and isn't easy. Google and other search engines won't index these sites well off the bat based on the domains and tailored but not_that_meaningful content. Maybe in the past that was true but certainly the SE ranking logic is smarter than that now.

The other area of trust would need to come from consumers. I know personally, if I was into brazilian funk, stumbled onto a machine-generated site like Sivers describes and find an album that I'm interested in buying, I would just look up that album at Amazon (or CDBaby) because I trust them. And that cycles back to the first point in that since places like CDBaby and Amazon have the authority, the index of brazilian funk results in the SEs are going to include these sites. So setting up these niche sites might make sense for sites w/ authority already, but if they already have authority why bother?

Categorizing is the wrong way to go, because different people categorize different ways. Assuming you have enough users, it seems to make more sense to go with some kind of bayesian recommendation system, i.e. "People who liked this product also liked..." That way you conform to the implicit categories people already hold, rather than force-mapping a blob of messy, overlapping categories into arbitrary, artificial divisions.

We do exactly this at work. The great benefit which is not mentioned in the article is SEO - Google will love you for it.

The problem with this approach is that most people don't try new music because they're into, say, Brazilian chill. There's a need for a trusted reference -- a friend's suggestion, a review piques your interest, whatever. So just setting up a domain for people who like that genre isn't necessarily going to go far. Most of the music would probably be crap, and not worth my time -- hence that need for reference. I won't assume that a domain name means good music, but I do, for instance, take Anji Bee's (http://www.anjibee.com/category/podcast/chillcast/) suggestions because she has a proven track record of finding and playing excellent chill.