- Developer Hub
- SparkPost API
- Email Tools
- Slack Channel
- User Guides & Migrations
- Submit a Ticket
- SparkPost Academy
- Deliverability Guide
- Email Explained
- White Papers & Guides
- Webinars & Videos
- SparkPost vs. SendGrid
- Contact Us
Artificial Intelligence and Your Inbox
Once upon a time, when I heard the phrase ‘AI’, I would immediately think of SkyNet, the self-aware computer determined to eradicate humanity in Terminator or the bio-engineered androids that were virtually identical to humans in the movie Blade Runner.
Flash forward several years, and AI is no longer something out of a science fiction story. Now, Artificial Intelligence is changing the way we do business. From workflow management tools to trend predictions, to even how brands purchase advertising (Google Adwords, PPC, etc.), AI is paving the way for innovation.
Consequently, it seems only fitting that AI is now helping to change the email marketing landscape.
Consider some recent statistics by the Radicati Group regarding email usage:
3.7 billion – the estimated number of email users worldwide.
269 billion – the estimated number of emails sent per day
2.4 million – the number of emails sent every second
74 trillion – the number of emails sent every year
49.7% – the percentage of emails that are spam
These figures are staggering. And while email still provides companies with the most bang for their marketing buck, this does raise a few questions. How do we, as marketers, break through all of the email noise and ensure we are communicating the right message to the right people, and at the right time?
Unfortunately with the economies of scale, many companies rely on semi-personalized email communications and batch and blast techniques, which unfortunately all but guarantees that their email ends of being part of the 135 billion spam emails sent each day. How can you make sure your organization’s emails make it through to the inbox?
Enter Artificial Intelligence.
If you think about all the applications that are now implementing and benefiting from AI, using it to enhance marketing and transaction email really is a no-brainer. However, there still seem to be some hurdles with organizations truly bracing the power of AI.
While the concept of AI and Email Marketing has been on the horizon for some time, many organizations while seeing it’s value are still relegating it to ‘back seat’ by essentially bolting it onto whatever legacy marketing platform they’ve implemented and grown accustomed to. While it certainly helps, they aren’t realizing the value that AI has to offer.
In order to see true success, AI must be front and center – driving marketing forward, not treated as an after-thought.
Please join us on August 22nd for a joint webinar with our partner BlueShift where we will not only discuss how to use AI to strengthen your email marketing efforts, but provide you with a real-life success story presented by LendingTree who is realizing tremendous benefits by being on the cutting edge of AI in marketing.
Title: AI in the Inbox – How to Put AI in the Driver’s Seat of Your Marketing
Date: August 22, 2017
Time: 11 AM PST/2 PM EST
Space is limited so Register Now
Sending email shouldn’t be that complicated, right? Let’s talk about a recipient on your list. In this case, we’ll call her “Jane.” Jane sees a piece of content you create and decides she wants to receive your newsletter. She signs up and even confirms her subscription (confirmed opt-in, score!). All of your future newsletters should land in her inbox, no questions asked. Seems simple, right?
I hate to break it to you, but that’s not how it works. There are many moving pieces to the email journey, and hitting “send” is the easy part.
Malicious senders in the form of spammers, phishers, spoofers and [enter other bad-guy types here]—including overly aggressive marketers—have turned the email journey to the inbox into a match of American Gladiator. Walls, flying tennis ball artillery, you name it.
To give you a better idea, I’ve outlined a high-level view of what happens when you send mail. While the particulars will vary based on the networks involved, it goes a little something like this:
1. Outbound spam filters (the “artillery”) The email passes through a spam filter before leaving the message transfer agent (MTA). Many email service providers (ESPs) have implemented some form of outbound filtering to protect against malware and phishing attacks that often originate from compromised accounts. Some providers even backup this type of machine policing with human review teams.
2. Blacklists (the “walls”) Before the receiving mail system will even think about accepting a message, it is checked against a number of internal and external IP-based blacklists (a.k.a. a DNS-based Blackhole List) to determine if the sending source is worthy of delivering mail to the mailbox provider. Blacklists like Spamhaus, Barracuda, and SpamCop exist to reduce the deluge of spam.
3. Internal content filters (just when you thought you made it!) Next, the email passes through a content filter on the receiving mail system that is especially designed to smoke out any bad links or attachments, in real time. If any of this content is deemed malicious, your message likely will be rejected entirely.
4. Commercial content filters (yep, more filters) At this point, the email passes through commercial content filters to see if it contains anything left unidentified by earlier gates in the system. You might be surprised at how many of these commercial filters are being used by major mailbox providers. In most cases, failing this step results in a message being filed as spam, rather than rejected outright.
5. Black-box filters (almost there…) Finally, the email is checked against various forms of other filters, especially at providers that offer mailbox-level filtering, to determine if the message should be placed in the inbox or spam folder. And if you’re sending to Gmail, there are also tab placement options besides only hitting the inbox.
And if all of this wasn’t enough, the logic supporting every point in the above list is constantly evolving.
Email deliverability is the term coined for measuring the success of a message reaching its intended recipient. Deliverability, as you can now imagine, is often perceived as a mystical (and sometimes unobtainable) thing because of the complexity involved. You might feel the same way when your mail is placed in Jane’s spam folder or rejected entirely.
The reality is that every serious sender needs the strategy and the tools to successfully deliver mail. Senders need visibility and the actionable data to continuously improve their deliverability.
If you’re curious and want to dive into details about the email journey your mails are taking, 250ok and SparkPost have partnered to provide SparkPost customers with some great tools for analyzing your own inbox performance.
About the Author:
Greg Kraios is a hardcore email nerd and the Founder/CEO of 250ok, the preferred choice for email analytics tools. Before 250ok, Greg spent several years at Salesforce Marketing Cloud (formerly ExactTarget) serving as their ISP Relations Manager, as well as providing deliverability consulting services to a variety senders including Angie’s List, Aprimo and PopularMedia (acquired by StrongView).
Let me start by saying I’m no mathematician, nor am I statistician, nor do I play one on TV. So occasionally, I need to look up the definitions of mean and median (cheat sheet: they’re the arithmetic average and the middle point in a series). I don’t think I’m alone in occasionally mixing up these closely-related (but nonetheless distinct) concepts.
Similarly, email deliverability, acceptance rate, seed list testing, and panel data are different ways of measuring delivery of email to the inbox. These metrics are related to one another, but they don’t mean the same thing. So, what exactly do they mean? And how can email marketers use these different approaches to evaluate the success of their campaigns?
Deliverability is the broadest of these terms. Deliverability is a fundamental metric for email marketers and other senders, because there’s no chance for a recipient to open, read, and respond to an offer if the email never arrives. Full stop.
Moreover, deliverability seems straightforward. If I send 1000 emails, and my server’s log files say 900 of those were accepted by receiving systems, then my deliverability is 90%. Simple, right? Yes… but no. What this figure represents is really the message acceptance rate. That’s because all the server knows is that the receiving system took the message, but not what was done with it. Did it go to the inbox? The spam folder? The sender really doesn’t know, because in both cases, the SMTP transaction is logged as as successful “250 OK”—SMTP itself doesn’t differentiate spam from legitimate email. That ambiguity is why server-side measures of message acceptance are just a starting point.
Acceptance rates don’t say anything directly about inbox placement. And inbox placement is what we really care about, for the simple reason that that’s overwhelmingly the most likely place a recipient is actually going to read our email. It’s safe to say that recipients who take the time to hunt through the various unsavory contents of a spam folder for a legitimate marketing offer are few and far between.
Addressing the question of inbox placement is why companies like 250OK, Return Path, and IBM Email Optimization offer seed list testing services. Here’s how it works: a sender includes special “seed” (test) email addresses at various ISPs among the recipients of their campaign. The seed list service providers then monitor those accounts with tools that determine where your email landed in their seed account—the inbox, the spam folder,… or perhaps if didn’t arrive at all. Because seed lists employ a known (and relatively small) set of addresses to test, they can give an answer about email performance of a particular campaign to those specific addresses. However, the information learned from seed lists should be considered at best directional. They can’t give you a comprehensive assessment of performance.
That’s where one more way to measure deliverability performance really becomes important: panel data. While seed lists use definitive results for specific email messages at a small number of addresses to extrapolate overall campaign performance, panels take something of the reverse approach. Panel providers like eDataSource monitor millions of real-world recipient inboxes (the owners of said mailboxes have agreed to participate in the research, by the way!) and aggregate data about message characteristics and performance over time. Thus, while seed lists are good leading indicators of the efficacy of a particular campaign, panel data is best for assessing broad slices of real-world performance. This includes overall message volume by sender, the behavior of senders in responding to bounces and feedback loops, and aggregate inbox placement across campaigns and time. Panel data is a powerful way to take in the big picture and to compare senders.
In short: email deliverability is a broad concept. It can be measured in multiple ways, including reporting message acceptance rates, performance to seed lists, and aggregate behavior as measured by panel data. All three methods of testing and measuring deliverability can be useful. All three also have their limitations.
When I consulted for email senders, I used to advise my clients to use a mixture of the three. Acceptance rates are the most blunt measure of how systems are technically working, but they don’t say much of anything about message performance per se. On the other hand, while seed lists provide a quantifiable way to see how a particular campaign is proceeding (and how specific tweaks to content, templates, etc., affect performance), they don’t give much of a big picture view. And panel data analysis gives you a good insight to how you’re doing relative to your industry and provides longitudinal benchmarks across campaigns over time.
(In fact, we use a combination of all of these methods to monitor different aspects of our own email deliverability performance at SparkPost. Empirical evidence is something we’re proud to stand behind when we talk about SparkPost’s inbox placement rate.)
However you slice it, measurement, testing, benchmarking, and tracking inbox performance are critical to the success of every sender. Getting a message to the inbox is just the beginning of your customer conversation—but it’s the only way it can get started.SparkPost © 2018 All Rights Reserved