Depends upon the mode. Either physical or delayed logical.

But the more highly available modes where the logs are written synchronously, typically don't perform well. Hence not used or are disabled after observing performance. Then your back down to sending archived logs. Thus data loss of up to the capacity of your online redo logs. And possibly then some, since you may still be in process of moving the last archive log at the point of failure.

Data guard, is data protection. If the loss is acceptable then go for it.

It all boils down to how much data loss is acceptable. If its zero (0) then your basically stuck with synchronous writing of the redo log and poor performance. If some loss is acceptable then you have some options.