This paper considers a random search model where some locations provide sellers with better chances of meeting many buyers than other locations (for example popular shopping streets or the first page of a search engine). When sellers are heterogeneous in terms of the quality of their product and/or the probability that a given buyer likes their product, it is desirable that sellers of high-quality niche products sort into the best locations. We show that this does not always happen in a decentralized market. Finally, we allow for endogenous location distributions and show that more trades are realized when locations are similar (in which case the aggregate matching function is urn-ball) but that quality weighted trade can be higher when locations are heterogeneous.