When you evaluate vendors for a translation project you should think of the cost as an investment with a typical risk vs. reward trade-off. High risk projects like the software or documentation for a defibrillator would require a low risk solution. A low risk solution would obviously be to choose a properly vetted translation firm with a proven track record, years of experience and a thorough QA process.
A low risk project might be translating an e-mail from your pen pal in Brazil. In that case a high risk solution like machine translation (Google Translate, Babelfish, etc) might be more than adequate.
When deciding who should handle your next translation project you should consider how much risk are you willing to take on a poor translation?
Do I even need to say that free has an infinite amount of risk?