Machine vs. Human

Our delivery team dives into a wide range of issues for a wide range of clients. But, the investigation always starts the same way. We want to identify when the delivery issues started and why. Then, we try to target what changed around that time. Was it the delivery statistics (opens, clicks, bounces, etc), the practices of the sender, or a combination of both. The results will usually point to something within two key parties that the sender is not currently pleasing: The machines that determine the delivery of a message, or the human viewing the inbox.

Sender and The Human

As humans, we’re all probably more familiar with what it’s like to review email messages from this perspective than we are to act like one of the thousands of filtering algorithms (yes, thousands) that major mailbox providers apply to all message delivery decisions. We’ll explore that relationship first.

How Are Expectations Set?

Does the user know what will arrive in their inbox after they provide their email address? Do they know how often it will come through?

Make sure these are clear and also that the same brand present at this point of collection is the same brand present in the upcoming messages.

How Are Expectations Met?

Does content sent to the user match what was disclosed during the collection of the email address?

Make sure the subject line intent and focus also matches the body content and intent as any break in the user experience from signup, to subject line, to content can all add to increases in spam complaints.

Do You React To The Feedback You Get?

When is a user removed from your list? Make sure that “unsubscribe” and “Mark as Spam” are not the only indicators you use to remove an email address.

Make sure to also pay attention to engagement and segment your list of recipients based on that information (i.e. send different content and at different frequencies based on how the users are interacting with your messages).

Sender and The Machine

For the more complicated topic of understanding how machines decide to deliver a message, it is best to remember that they are mainly reacting to how humans have previously reacted to messages from the same sender. I say “mostly” because there are some universal mistakes to avoid.

What Elements Are Included?

By this we are referring to the “content” included in the message and the “header” information seen behind every message. As we discussed in another recent post, Email’s A Stage: A Look Behind the Curtain of Your Email Program, these are the main two areas where the sending elements can live; the headers “behind the curtain” and the subject and content “in front of the curtain.” Be very intentional with all that is included in these areas. All pieces of an email message can affect its delivery.

Is Your Reputation Being Diluted?

Are the URIs, image hosting domains, IPs, etc. seen elsewhere within any other email messages? Remember that filters are getting more sophisticated and are fingerprinting a lot of elements of a message. Try not to share any elements that are connected with a lower sending reputation.

Has A Proper Warm Up Concluded?

I know…now i’m just rhyming for fun. But, this is meant to bring focus to the necessity of a proper warm up for all mail streams. We’ve previously weighed in on the exact details of a proper warm up plan. Making this a priority comes from remembering that, in the eyes of mailbox providers, most marketing mail is “guilty until proven innocent.”

When a new sending IP or sending domain is used in a mail stream, the mailbox filters don’t know how the “human” has reacted to that mail yet, so they will default to protecting the user and possibly not delivering all initial messages to the inbox. If you have data that points to the recipients who are most likely to engage in the initial campaigns, you will help to secure a positive sending reputation for a new mail stream by sending messages to those inboxes first.

This balance of “machine vs. human” is definitely akin to the “chicken and the egg” debate. And, like all email delivery strategies, it takes constant attention and a scientific approach including various A/B tests to confirm the right strategy for each specific sender.

To test your knowledge of email delivery do’s and don’ts, check out our interactive Deliverability Guide.

Happy Sending!



Jacob Hansen
Jacob comes from a background in technical account management and delivery analysis for the last six years, and has been with SendGrid's Deliverability Consultant team for the last two years. He enjoys spreading knowledge to help the email community send more "wanted email" and to help senders realize their full potential. Originally from Nebraska, but living in Colorado long enough for it to feel like home, Jacob enjoys a lot of what the Denver restaurant, bar, brewery and music scenes have to offer.