[pbhj]

No you couldn't, suppose your A does better in the first market and your B worse (accounting for size) in the second. You choose A. Which does even worse in second market. B was doing the best job for the second market, it is just that market is less open to your product, B would have totally pwned in the first market.

You have to do the A/B within the same market to get useful results.

[philwelch]

Use a good enough mix of similar markets and you'll get not-entirely bogus results. The problem you're talking about had occurred to me too, actually, since markets vary widely in climate. Market it in Alaska and people never, not even in the summer, get used to light clothing for long enough to prefer blankets over heavy clothes when on the couch anyway. Market it in Florida and it never gets cold enough to use a blanket. You'd have to control for climate and culture, which across the US would be challenging (to say nothing of the world).

There are ways to do this in the same market though. Something like Vibram FiveFingers is viral enough that you could probably seed like five Vibram wearers in a single market and have each of them refer to the product under a different name. Then track your incoming search queries or orders and match them by market. If more people order "FiveFingers" than "Toe Shoes" than "Foot Gloves" you have yourself a product name.

Well, until someone clever on the internet exposes you. Then you get free publicity as that company that sells the weird shoes and can't decide what to call them.