Study Finds Google Mail Improving at Spam Recognition

Posted on October 2, 2006

The New York Times reports that a study from Lyris has found that Google's Gmail service is getting much better at recognizing spam. Google improved 15% from Q1 2006 to Q2 2006.

In its latest quarterly test, Lyris tracked more than 57,000 e-mail messages, sent from 57 different businesses and nonprofit organizations to scores of e-mail addresses it owns in many different domains. The messages included marketing pitches for electronics and perfume, and noncommercial matter like a wine newsletter. None of the senders were Lyris clients; all the recipients had signed up to receive the messages.
In the most recent test, 3.3 percent of those e-mail messages were treated as spam by American Internet providers — something that marketers call a false positive.
Since the distinction between spam and legitimate e-mail is not always readily apparent, most providers have had to learn which marketers are legitimate. False positives result when they get it wrong. Individual marketers also get more false positives when they send badly aimed pitches, leading recipients to tag their messages as spam.

Spam recognition by email software programs must continue to improve. False positives with mail from marketers may be dissapointing to retailers but false positives involving personal mail are unacceptable. A 3.3% false positive rate is still troubling.