Designing around spam filters
A guide on email newsletter design wouldn’t be complete without a section on spam
filters. You have to design your emails a certain way, if you want your messages to make
it past them. But you can’t go overboard, either. Spam filters can smell fear. If you try
too hard, they know it, and they attack. So just be cool, act normal, and don’t make any
obvious mistakes.
How anti-spam systems work
Before getting into the nitty-gritty design tips, you need to know how all the various
anti-spam mechanisms work out there. Once you get an overall understanding of how
they work, designing around them will be much easier.
Bayesian Filtering
This is one of the most important ones to learn about, since it’s installed in so many
email applications these days (like Outlook, Mozilla Thunderbird, and Apple Mail).
Bayesian filters work by watching users classify email as “junk”
(such as when they click a “this is spam” button). It reads the junk
mail, compares it to other emails you called “junk,” and looks for
common traits in the subject line, the content, the hyperlinks, the
sender, etc. Over time, Bayesian filters learn to scan for those “traits”
in every email message you receive. Every time they find something
that looks “spammy,” they assign a “score” to it. For instance, using
“Click here!” might get you 0.7 points. Using bright red fonts might
get you 2 points. Including the word, “mortgage” might get you 1 point, but using the
word, “Viagra” will get you 5 points. Once an email exceeds some threshold (set by the
user), the email is classified as spam, and thrown into the junk folder. It’s amazing how
many different things Bayesian filters look for. To read some, go to the MailChimp Blog
and search for “Funny Spam Assassin Criteria”
MailChimp also comes with a spam filter checker (look for the “Inbox Inspector” under
our Account tab). It’ll tell you exactly what your email’s spam score is, and what you
need to change if you want to improve your chances with spam filters.
Black Lists
A while back, some server admins got really, really, angry at all the junk mail they
received. So they started to track the IP addresses of the servers that sent them spam,
and put them on “blacklists.” Anytime they received email from a server on their
blacklist, it was deleted immediately. It worked pretty well. So they started to share their
lists with other server admins. And other admins started to add on to those blacklists.
Pretty soon, the blacklists got very, very big. Large ISPs started to sync up with them. If
you send email that’s very “spammy” you could end up on one of these blacklists
(whether the recipients are opt-in or not). Once you make it on to a blacklist, good luck
getting off. Geeks are notoriously difficult to negotiate with (we know, because we’re
geeks). This is why you need to make sure you only send to recipients who gave you
verifiable permission to email them, and you need to have proof that each one of them
opted in to your list (such as through a double opt-in system). Want to know the easiest
way to get blacklisted? Let your company’s sales team blast out an email newsletter to a
list of “prospects” that they collected from conferences, and that they scraped off of
websites. You’ll get blacklisted, guaranteed.
Email Firewalls
Managing email servers (and incoming spam, viruses, and phishing attacks) can be a lot
of work. So large corporations usually install “email firewalls” (Google the terms,
“Barracuda Firewall” or “Postini” for some examples) to handle their incoming email.
Think of them as “spam filters on top of spam filters.” They’re big, heavy-duty
gatekeepers, and they’re not friendly at all.
They often use a combination of Bayesian-style/adaptive filters, community reporting,
blackhole lists, and a little bit of proprietary “magic pixie dust” to keep spam out of the
company. Most of the time, when your email’s not getting through to a larger company,
it’s their firewall. You can think of these firewalls as kind of Xenophobic and paranoid.
They’re all twitchy, and tend to ask questions like, “Okay, is this sender new to me? Why
is he sending copies of the same, exact email, to a bunch of people in our company?
Spammers do that kinda stuff. Hmm, how long has their server been around? Can I
really trust this sender?” Spam firewalls are usually only a problem when you first start
sending campaigns to a big client or something. You’ll experience some deliverability
issues in the beginning, because you’re “new.” They’ll eventually “learn” to let you
through. To expedite things, you may have to ask the IT people in charge of the firewall
to “white list” your IP Addresses (or the IP address of your email service provider).
If you’re interested in learning more about email firewalls, search for “email firewalls” at
the MailChimp blog:
http://mailchimp.blogs.com
Challenge/Response Filters.
These are more common among “at-home” recipients (because they’re too intrusive to
use at work). When you send email to someone with a challenge/response filter, here’s
what happens. If you are not
already in that person’s “buddy
list” or “address book,” then
you’re considered a stranger to
him. And if you’re a stranger, you
could be a spammer. So their
challenge/response filter sends
you an automatic reply with a
question that you have to answer,
or some link you have to click
(this is to prove you’re a human,
and not a spambot). The screenshot you see to the right is an example of a typical
challenge/response reply, from Spamarrest, a very popular vendor.
The thing to remember here is that you have to be “white listed” if you want your emails
to get through. So when people fill out your opt-in forms on your website, ask them to
“please add our email address to your address book...” Use your opt-in process as a way
to “set expectations” and get “whitelisted” up front. Whenever you send a newsletter,
make sure the “reply-to” address that you use is valid, and that a human checks it after
each campaign. You can expect to receive a couple auto-replies like this after every
campaign you send. Whenever we send our MailChimp Monkeywrench newsletter, we get
about half a dozen of these autoreplies. I have to manually click each one of them if I
want the email to get delivered.
Labels: screenshot, spam, spam filters, spambot, vendor

