When is automated remote usability testing useful?

There seem to be a fair few services these days offering automated remote usability testing, so I'm assuming there are a fair few people using those services, which implies that they ARE useful - under the right circumstances.

But... what are those circumstances? When do automated remote tests yield data that usefully informs the design process?

Note: I'm not asking about moderated remote tests, with live interaction with the user - those seem close enough to face-to-face tests to be uncontroversially useful.