This study attempts to carry out meticulous investigation upon the translation errors by machine translator in Facebook. Having known that machine translator in Facebook often performs inaccurate translation, we tried to classify its translation errors in the first place. We focused on Bahasa Indonesia to English translation, and then providing logical explanation of the phenomena. The data of this study were collected from several users' posts on Facebook that are translated from Bahasa Indonesia into English. We consider the frequency and content analysis, the textual, grammatical and cohesiveness of translation. The result revealed 55 errors within 18 Facebook posts. Most errors were classified into the syntactic errors with 22 or 40.01% errors in total. Meanwhile, the mechanic errors were found 16 items or 29.09% and the semantic errors consist of 23.63%. The rest of the data were categorized as the cultural errors which consist of 4 issues or 7.27%. This study, afterwards, may help the readers understand the common translation mistakes that the robots produced. Therefore, the society will be more aware and the possibility of misunderstanding may decrease. Besides, it will also be helpful for future social media developers to consider more about the language features on their applications.