BTW, Cloudmark’s Spamnet which I’ve been using since the first beta version seems to have really nailed its algorithms with the latest beta version (Beta 9) of their MS Outlook plug-in. I’m now finding that over 95% of spam is being filtered correctly and the number of false positives seem to have dropped quite a bit.
It’s also worth noting that the false positives are pretty permission-based lists that have at least one of these three characteristics:
1. In frequent mailers — companies that don’t send for a long time seem to get picked up (probably because people don’t recognize them). CNMA is in this camp.
2. Low value lists — things that are probably of far less value than the subscriber would have expected. People “block” the messages rather an unsubscribing.
3. Drifting permission — mailers seem to be pushing the bounds of permission and getting penalized for it. For example content-heavy newsletters get through but if they send a “special offer” from a “valued partner” they get tagged as spam.