eConnect Email Help Center

Beginning with Outlook 2003, Microsoft offered a built-in spam filter in its mail client, but has provided little documentation as to how messages get filtered. A company called MAPILab reverse engineered the spam filter in Outlook 2003 and believes that its spam filter gives different weights to different emails depending on several categories. According to MAPILab, the following steps are taken by Outlook to consider whether email is junk or not:

• message sending time check--messages sent on a Saturday or Sunday are more closely correlated with spam than messages sent on a Wednesday

• check of the message subject for words in uppercase--if words in uppercase make 25% or more of the total number of words, then the email is more correlated with spam

• check of the sign number in the message subject--this test calculates the ratio of signs (symbols which are neither letters nor numbers) to the number of signs, letters, and numbers. If the ratio exceeds 8%, email is more closely correlated with spam

• check of duplicate character number--this test counts the maximum number of duplicate characters in the message subject. In many spam messages there is a sequence of meaningless (to recipients) symbols separated by several dozens of spaces

As with Norton AntiSpam, since Outlook does not provide reason information, we have installed our own artificial intelligence spam filter which approximates Outlook results and provides suggested keywords that cause messages to fail.

We will display up to 45 keywords with a weight greater than 0.6, in descending order from highest to lowest weight. Keywords that are weighted more highly are more closely correlated with keywords that cause failure.