Chris's Wiki :: blog/spam/MailerMotivations

Understanding the motivations of mail service vendors

October 5, 2011

From a comment on my entry about how modern mailing services should work:

Like any paid service, if they are not providing good service, like letting spam through, people will move to another one who does block spam better. This is their incentive to deal with spam.

This is a common misunderstanding of the incentive structure in operation.

This is not an incentive to deal with spam, this is an incentive to not get blocked. The two are very much not the same thing. Very few people who use a mailing service care very much about whether or not it sends spam; instead, they care about whether it can deliver their email mailouts (and do so without having them scored as spam by the recipients). No one is actually paying the mail service to not deliver spam, so it has no direct incentive to do so; if anything the straightforward incentives go the other way, towards giving spammers the benefit of the doubt so that they keep paying.

(There is some indirect incentive, in that people sort of care about your overall reputation and your overall reputation is affected if websites have lots of stories about how you are a big spam source and have a spam problem.)

For many mail service providers, trying to block spam is one of the most cost effective ways of keeping your mail delivery rates up. However, if you are big enough you become like GMail, Hotmail, and Google Groups; you are too big and popular for anti-spam systems to block unless you become a truly epic emitter of spam. At this point providers lose much of their business incentive to block and reduce spam.

(Note, however, that every mail service provider does less than they could to block spam through their service because every MSP knows that if they adopted full best practices, they would go out of business because customers would find them unattractive. This should not be surprising.)

Nor do mail service providers have a real incentive to refuse to do business with all spammers. Some spammers make unattractive customers (they are unsavory, a big support burden, may attract legal attention, and are high risks to cheat you on payment). However, some spammers make very attractive customers; they pay on time and well, they don't cause you support headaches, they're reputable businesses, and so on. They just send UBE and want you to send it for them. See many examples (and also).

You may appeal to moral qualities as part of the 'incentive' that providers have to deal with spam, ie that they really intrinsically care about not sending spam and are not merely doing it because of business needs. However, mailing service providers (and their employees) have consciously chosen to enter a line of business that draws spammers to it and intrinsically enables spam. It's hard to avoid the conclusion that they consider some degree of spam a necessary evil in order to make money; from there, to misuse an alleged Winston Churchill witticism, it's only a question of how much spam.

Comments on this page:

From 71.56.100.181 at 2011-10-05 11:44:31:

I am the commenter you are quoting, and I'm surprised at how thoroughly the point was missed. I thought it would be obvious.

To any email provider, the term "spam control" means filtering INCOMING spam from the Internet to prevent it from getting into customer's mailboxes. That's the only thing any email customer cares about, and is something that would motivate a customer to look for a new provider if their current one was not doing a good job.

A customer of an email service doesn't care if other customers are sending spam out to the Internet, unless, of course, the whole provider is blocked. However this is so obvious I can't imagine the need to write paragraphs about it.

By cks at 2011-10-05 11:51:51:

We're talking about different things, then. In both this and the previous entry I'm talking about providers of mail(ing) services, firms that exist to send email out for website operators and the like. These people generally do not have incoming email or customer mailboxes, and when they do (such as GMail and Google Groups) their mailbox customers do not really care about how much spam the mail server sends out, only about how much spam it accepts.

I thought this was pretty obvious, in part since my initial entry spent a bunch of time talking about services for sending email out, but I apparently wasn't clear enough.

From 78.35.25.22 at 2011-10-05 12:59:56:

It was obvious.

—Aristotle Pagaltzis

From 76.168.94.108 at 2011-10-05 14:19:12:

my name is John, and I work for SendGrid, an email infrastructure provider. We don't see fighting spam as a pure expense. On the contrary, the more effective we are at rooting out the pests, the more the big ISPs will listen to us. When a spammer gets email through our system, and the downstream ISP blacklists one of our IPs, it benefits us (and our clients) if we have a good reputation. We can fix the problem, ask for the blacklist to be lifted, and move on. That is, we fall into your "incentive to not get blocked" category.

Similarly, customer support could be "pure expense", but in our case we are happy when free users get phone support. We're spending money on users who are using our service and not paying us anything, with no commitment, because it generates goodwill. In general, enough people convert to paid accounts to make it a happy solution for all involved.

Vs your point that we have no "real incentive to refuse to do business with all spammers" -- we have several. 1) They decrease our reputation wrt downstream ISPs, as above. 2) Fraud -- they don't like to pay us.

Re: "Some spammers make very attractive customers; they pay on time and well, they don't cause you support headaches, they're reputable businesses, and so on." -- that's not my understanding. Our business model means that even well-paying spammers are not good clients. We can and do refuse to do business with high-volume companies if they have poor quality deliverability (ie: they scrape email addresses or otherwise don't prune newsletter lists), or get too many spam complaints.

Lastly, we allow and encourage senders to have dedicated IPs for exactly the reasons you're stating. Its kind of middle ground for what you're after versus making life easier for customers. We send email via the user's dedicated IP, which means if they send spam then they get the problems -- they have immediate incentive to keep their email stream clean.

- John (john.mitchell@sendgrid.com)

ps: if this feels too much like an advertisement, feel free to edit/delete. I just wanted to say "hi" and provide an alternative data point or three.

Written on 05 October 2011.

«	On CPython, cell objects, and closures

Thinking about event loops versus threads