You can probably guess how reviewers are rated. Did you provide a good review (as in, read the paper and provide constructive comments)? Did you provide a review on time? Did you write a biased review (a giveaway would be to say the paper's acceptable if and only if the author cites your papers)? If you wrote a good review then you get a high score, otherwise you get a low one. If you get a high score then editors are theoretically more inclined to invite you in the future; the converse is true if you get a low one.
Whether or not ratings are shared between different journals of the same publisher depends on the publisher and the editorial boards of both journals. Obviously, journals can only share reviewers if they have some overlap in scope. There're also technical limitations to linking reviewers. Different journals use different systems, and while it's theoretically possible to link all of them to the same database, it's not always done. Beyond that the editorial boards must also agree. If an editor is relying on his or her personal network, (s)he might be less willing to share. Comparatively, if an editor is using Google Scholar, Web of Science, etc to find reviewers then (s)he might not object.