Why does some spam get missed? Why do some messsages get marked incorrectly?
It is not possible to determine, based on a single email message, whether that message is legitimate with 100% confidence. For example, the fact is that most spam messages are advertisements, and much of spam scanning consists of looking for messages that look like advertisements. However, marking everything that looks vaguely like an advertisement will also mark every message about a sale or special offer about which users have requested to be informed. We call this incorrect marking of a legitimate message a “false positive”. Any system that is sensitive enough to mark every single spam message will also end up with many, many false positives. Likewise, any system permissive enough to never mark a legitimate message will miss a great deal of spam. We’ve tried to strike a balance, keeping the number of false positives extremely low, while still trying to mark the vast majority of spam. In the pilot program for our spam scanning system, the vast majority of false positives were solicit
