"A partner of ours periodically sends an email to hundreds of reporters from their personal mail account. Partially in hopes of reducing their spam quotient, they send the message to 50 addresses at a time. My question: does sending the message several times to limited batches of addresses affect their chances of being labeled a spammer, or their message being labeled as spam? Thanks for any wisdom you can provide.
Alan"

NPA: Alan, thanks for a good question. The answer is, as economists say, yes and no. That's because spam is filtered in two places.

Yes, the spam filters that ISPs use to determine spam take into effect the number of emails that come into that ISP at one time. So if 500, or 5000, (or maybe only 2 identical emails, who knows?) come into AOL at one time, AOL notices and may deal with them differently.

However, the spam filter that operates on a recipient's own computer does not know how many other people got that email, so it determines it to be spam or not depending on the content of the email and the header. Here's an example: My wife sent me an email yesterday with just the URL we had been discussing. Emails whose sole content is a link get caught in my Norton Anti-Spam, so hers was labeled as spam, even though there was just one email, and even though it came from another email address on my domain! Note: an ISP may pass along some value judgment to the recipient. AOL and Yahoo! are examples of ISPs that put emails into different folders based on values they identify.

Another contributor offered this suggestion:

It depends on both the sending and receiving Mail Transfer Agent (MTA - what most of us call an email program). Some MTAs use the old sendmail method to deliver messages, i.e.:

1. group recipients by domain
2. connect to a domain once and deliver the message once to all recipients at that domain
3. disconnect
4. repeat 2 and 3 until all domains are served
5. report on success/failure

A more deliberately designed MTA like qmail, does it a little differently:

1. connect to the domain of the first (or nth) recipient
2. deliver the message to that recipient
3. disconnect
4. report success/failure
5. repeat until all emails have been sent

Now, it seems obvious that the first method would be faster, but it isn't. You can trust me on this, or see http://www.lifewithqmail.org/lwq.html#multi-rcpt for a bit of explanation. Suffice it to say, the 2nd actually works better.

For our purposes today though, my point is simply this: If your partner delivers messages via method one (Multi-RCPT), then batching the delivery does make a visible difference to the average receiving mail server. On the other hand, if your partner's MTA uses the second method (Single RCPT), the only thing that batching the emails succeeds at doing is using a lot more system resources than necessary.

In terms of not being dubbed a spammer, there are LOTS of options. Bonding is probably the most reliable, but running your own messages through Bayesian filtering systems like SpamAssassin will give you clues to help construct a message that will pass someone else's.

For more discussion on ways to increase the deliverability of email, please see our blog's discussion of the Certified Email system being planned with GoodMail, Yahoo! and AOL.

Also, as a reader reminded us, the Direct Marketing Association has adopted guidelines for email authentication which are mandatory for members and certainly recommended for all other email marketers. You can download the 10-page pdf here: http://www.the-dma.org/antispam/E-MailAuthenticationComplianceFINAL.pdf

March 2006