You don't have to scan the receiver's mail or use the sender's reputation. So the system can be bootstrapped without analyzing or rating any messages at all. All you do is assign an initial ham,spam count for active users that is higher than inactive users. The system then self-adjusts the statistics as mail flow comes in.

Using the above bootstrapping technique, with only 1M messages, we can achieve 99.69% catch rate with zero false positives which was measured on a 130,000 message test set.

It is quite amazing to see.

Totally content and language independent.