Marydee Ojala, Editor of ONLINE looks at Scopus, World of Science and Google Scholar.

It seems to me you can compare subject coverage, date ranges, interface design, search functionality, number of journals, type of format (citation, abstract, full text), number of hits resulting from a search, and cost. What you can't systematically determine is quality of results. For that, we only have anecdotal evidence. Here's one small anecdote.

Ojala decides to search for herself in both Scopus and WoS and then uses Google Scholar to verify and look at the articles that cited her.