Starting in one very tight niche (e.g., as you suggest, one geographic area) is a good idea. But I’d also add another very important point:

Build one side of the marketplace first, by providing something of value (other than the marketplace itself, which won’t exist yet) to one side of the marketplace.

For example, if you want to build a marketplace matching taxis to users, build the network of taxis by providing a useful tool to taxi drivers to allow them to monitor their mileage / fares / etc. (which is what Hailo did). You can get hundreds or thousands of taxi drivers on your “network” (=using your tool) before you even mention the marketplace idea. Then when you are ready to launch to users, you already have your supply side in place.

Or provide a portfolio service for designers, so you get lots of designers hosting their work using your service. Then offer to match people who need designers with the designers on your network who are looking for more work.

This approach is covered in more detail in this Pandodaily post from 2012:

http://pando.com/2012/11/20/liquidity-hacking-how-to-build-a-two-sided-marketplace/