Last time we talked about the words Spam and how to avoid them. It is good to know and be able to predict the consequences for your email marketing campaigns, but how does it work in the end?
The CAN-SPAM law regulates commercial email, that is, any email sent by a company that tries to promote or sell a product or service.
The CAN-SPAM rules are as follows:
- Do not use false or misleading information.
- Do not use a misleading subject line.
- Tell subscribers where you are.
- Offer subscribers an easy way to cancel their emails.
- Respect all exclusion options quickly.
Penalties for not following the law can add up to $16.000 per email, so it is important to follow the rules. For more information about the CAN-SPAM law, visit the Federal Trade Commission website.
There are different algorithms and programs that scan the content of the email and contrast it with a determined list of words. Thanks to machine learning they can understand if the word is used to trick (for example, the email is full of phrases like “100% more”, “Become a member”, “No credit check”, etc) or it is just a coincidence that is there.
In most cases it refers to bayesian spam filtering, a technique that usually gives low false positives in spam detection. It is one of the oldest forms of spam filtering, with roots in the 1990s.
A bit of history of spam
Spam (from the contraction of “Spiced Ham“) was the food of Soviet and British soldiers in World War II. Later, the British comedy group Monty Python began to make fun of canned meat because of its poor quality and extensive use. The typical menu of a cafeteria at that time contained this product to replace bacon and other more expensive types of meat. It is worth seeing the sketch, it is only 2 minutes.
Years later, with the growth and free reach of the Internet, some new users were mistakenly sending personal messages to a whole list of contacts or discussion groups, which generated annoyance and loss of time (and even money) for the other users who received this irrelevant and undesired traffic. In 1993, someone compared such unwanted intrusions to spam: the messages were like canned meat, nothing could be eaten without encountering a stiff.
The most “tricky” thing is that spammers usually get their email addresses from legal sources. In many cases users leave their data everywhere, accepting the terms and conditions without reading them, subscribing to promotions, or simply forwarding the emails without hiding the addresses in copy.
Some of the main sources of addresses to send the spam are:
- The websites themselves, which often contain the address of their creator, or of their visitors (on forums, blogs, etc.).
- Usenet newsgroups, whose messages often include the sender’s address.
- Mailing lists: you only need to sign up and write down the addresses of your users.
- Emails with jokes, chains, etc. which internet users usually forward without hiding the addresses, and which can accumulate dozens of addresses in the body of the message, and can be captured by a Trojan horse or, more rarely, by a malicious user.
- Pages that request your email address (or your friends address to send them a page of an email) to access to a determined service or download.
- WHOIS databases.
- Illegal entry into servers.
- By trial and errors: addresses are randomly generated and then checked if the messages have arrived. A common method is to make a list of domains, and add common “prefixes” to them. For example, for the domain wikipedia.org, try [email protected], [email protected], [email protected], etc.
In addition, it is common for the sender of spam to monitor the reading of the emails to find out which addresses work and which do not. This is done with famous web bugs or pixels -they are small 1×1 pixels images in the HTML code of the message. This way, every time someone reads the email, your computer requests the image form the sending server, which automatically records the request.
Firewalls and spam detection systems
Well, these (pixels) are one of the signals to firewalls that mean the email may contain junk information.
Other signs for algorithms can be:
- Excessive use of capital letters or characters.
- Exclamation marks and questions (?,!), especially if they are used a lot in a row.
- Colourful fonts in different sizes.
- Errors in the HTML code.
But it is not so strict today thanks to Machine learning. For example, platforms such Gmail or Hotmail attach great importance to the historical relationship between a sender and a receiver in order to consider whether an email is spam or not. In other words, if they have not had contact with the person to which they are sending the email before, it is more likely that the message will not pass the filter if it contains a suspicious word. And if you send an email to a frequent recipient, it does not matter if you include the whole “bad” expressions list , your message will still arrive well.
Reputation in email marketing is like a parent’s trust in their teenager: it is hard to gain and easy to lose in a single moment.
Email providers take into account many metrics to determine the reputation of their senders, including spam complaints, how many unknown users you contact by email, whether you are on an industry blacklist, and more. All of this is reflected in a number between 0 and 100 which is called Sender Score.
- 0-70: VERY BAD. You have to improve your score as soon as possible, take care of your domain, warm it up, do not send invalid emails and verify the DMARC, DKIM and SPF indicators.
- 70-80: The worst is behind, but keep improving and perfecting your campaigns.
- 80-100: congratulations =))
For example, in long term is very important that the prospects:
- Open your emails.
- Do not mark them as spam.
- Do not delete them without reading.
- Add your address to their contact book.
They are one of the most likely causes of low metrics such as deliverability, open rate or CTR, or they have decreased dramatically campaign after campaign.
Spamtraps or “trap emails”, are the ones created not for communication, but to attract unwanted mail.
There are two clearly differentiated types.
- Pure spamtraps. They are specially created to detect pure spammer, which is the one who sends to mailing lists that are created by themselves. These lists are composed of addresses that have never been published and that’s why this type of spammers are the most dangerous, since they may even have entered our database via crawling or trojan.
- Recycled spamtraps. The second type is, as its name indicates, email accounts that have been previously used and were abandoned by their users (for changing their job, etc). The mail provider here identifies spammers and uses them in a curious way of two phases:
- In the first phase these accounts will report a hard bounce, and this will be reflected in the statistics of the campaigns. Hard bounce means that the account or the domain does not exist or that the provider has blocked the delivery.
- These accounts are then formally activated as spamtraps and signal those senders who do not clean and manage their databases.
Now, if you want to review the list of the SPAM words, here we published it a few months ago and since then we have kept it updated.