In 1971, the world’s first email was sent. We’re guessing that the first spam message was sent shortly after.
Today, the world of email (and the spam that plagues it) is way more sophisticated. From 1 million email users in 1995 to 2.5 billion in 2015, email usage has exploded. And now mailbox providers and spammers alike rely on artificial intelligence to do battle.
Your average mailbox provider like Gmail uses a ton of AI to determine if emails are spam, then delivers them accordingly. But that’s just the beginning.
Now, mailbox providers use advanced AI and machine learning to not just stop spam, but to significantly improve the email user experience.
You don’t always see this AI at work. But boy does it affect your email marketing campaigns.
We’ve all felt that horrible sinking feeling when we realize our latest, greatest campaign is going right to spam.
Or the misery of getting buried in a prospect’s promotions folder under the communications from a bunch of other brands.
Artificial intelligence systems make these decisions. They use very specific rules that ingest lots of data from millions of users to determine how emails should be classified. These machine learning models then improve over time based on user behavior.
That’s the reason emails stop appearing in your Gmail inbox if you mark them as spam. It’s also why Gmail stops delivering emails to your primary inbox if you don’t open or engage with a certain type of email or email address. In both cases, and many others, Google’s algorithms learn from your behavior and adjust accordingly.
Email clients are constantly assessing factors like text classification, your interaction with emails, sender reputation, and thousands (if not millions) of other factors to make these decisions in real time.
The use of AI in the email industry isn’t new. In fact, AI has been used since the machine learning driven Bayes classifiers of the 1990s to filter spam. From there, it evolved to focus on reputation filtering, looking at your send patterns and finding abnormalities. Today, the AI systems used by mailbox providers also look at how people at a subscriber level interact with your communications.
This all presents a real challenge for marketers.
Nobody knows exactly how different mailbox providers use AI and machine learning to filter your emails. There’s no list lying around of all the rules Gmail or Yahoo or any other provider uses to determine where your email ends up. Each provider uses different criteria, tests, and models to inform email decisions. And these decisions are driven by each company’s unique business opportunities and challenges.
On one hand, this means email marketers must do what they’re supposed to be doing in the first place:
Send great emails that users love to open, read, click on, and engage with.
If you do that, you’re well on your way to having users engage with your emails, which in turn signals to back-end AI algorithms that your emails are trusted and worth prioritizing.
But on the other hand, this means the stakes have never been higher.
The real danger isn’t that you send an email your customers ignore. It’s that you send too many emails that users ignore, mark as spam, or move out of their primary inbox. Over time, the robot overlords that determine how your emails get delivered are going to learn one thing and one thing only from these user behaviors:
Your emails aren’t as trustworthy or valuable as other emails hitting a user’s inbox.
The user is in total control. And every action the user takes informs the machine models that determine how your future emails are delivered.
Unfortunately, there are too many variables involved for any one human email marketer to grasp and act on here.
The AI models themselves run on hundreds and thousands of machine learning rules. These rules are informed by many user behaviors and actions.
At Return Path, we help businesses communicate more reliably, effectively, and securely. We’ll be writing about how email marketers must respond to the changes brought about by AI in email. But the first, and best, place to start is by benchmarking how your emails perform with actual users.
You’ll want to use human experts or machine tools to gauge your current deliverability and engagement rates. You’ll also want to determine your sender reputation, which informs how AI models treat your emails.
To do that, take Return Path’s comprehensive reputation measurement assessment here.
Lauren McCombs has been at Return Path for 3.5 years and currently manages the Data Science team.