Study: Spam Filters Often Lose E-Mails

AP – As spam-fighting tools become increasingly aggressive, e-mail recipients risk losing newsletters and promotions they’ve requested. [Yahoo! News – Technology – Spam]

A new study attempts to quantify missed bulk mailings. Return Path, a company that monitors e-mail performance for online marketers, found that nearly 19 percent of e-mail sent by its customers never reached the inboxes of intended recipients.

The figure, for the last half of 2003, is up 3.7 percentage points from the same period in 2002.

In some cases, the messages weren’t delivered at all; in other cases, messages wound up in spam folders that are rarely checked. Though technical glitches can also cause mail to disappear, Return Path blames most of the deletions on spam filters.

3 comments for “Study: Spam Filters Often Lose E-Mails

  1. 3/15/2004 at 11:00 am

    Personally, I’m finding that the best indicator on whether inbound email is spam or not is whether it adheres to technical standards (does it have valid headers; is it sent through mailservers known to have sloppy enforcement of relaying? are the timestamps approximately accurate (mail from the “future” and “past” regularly arrive and is generally much more like ly to be spam)…

    I use Spam Assassin [1] as my only tool at the moment, and I find that it (including its auto-learning Bayesian filtering) is reasonably accurate. I’m also considering Greylisting [2] but haven’t gotten there yet…

    Commercial newsletters I subscribe to usually get through; if they are technically deviant (using odd HTML colours, containing HTML only etc) I can always whitelist the senders if I really want their mail to come through (but if they can’t even bother sending technically correct/valid emails, I’m not always sure I can be bothered…)

    [1] http://www.jacobsen.no/anders/blog/archives/2002/10/16/spam_hype_and_spamassassin.html
    [2] http://www.jacobsen.no/anders/blog/archives/2004/02/09/greylisting_to_kill_spam.html

Leave a Reply

%d bloggers like this: