Spam trial results (round 1)
I recently posted about a trial we were doing to test a new spam checking system. I promised those that participated that I'd update this blog with the details once they were in.
Well there's some good and bad news. The good news is that the trial went well, and thanks to the people that helped out, we were able to gather some great and accurate data. The bad news is that the system we were testing didn't work as well as we expected in the first round of tests. It was good, but it turns out that our existing SpamAssassin system with our custom rules, address book whitelisting, and personal bayes DB is actually amazingly good.
However there is some work we'd like to try with the new system, so I'm planning on making some changes to the test setup, and will be running a further trial next week. I'll be emailing all the people that were involved with the previous trial about this, so that they can hopefully help out again. The nice thing is that the things we ask the people helping us out to do (reporting all spam and non-spam accurately), actually helps their spam filtering as well, so it's a win/win situation for everyone.
More details early next week!