Data-driven Insights and Opportunities in Email
Peter Wilson - CEO, LashBack
Mailcon – Las Vegas
January 2018
LashBack’s President and CEO, Peter Wilson, recently had the honor of speaking at Mailcon in Las Vegas, Nevada. His presentation at the event focused on message quality and various characteristics of an email that can affect inboxing rates among different CEPs.
“Email is a powerful but complex marketing channel,” Wilson says in the presentation. “In order to succeed, you need to do the right analytics… at the right time, using the right data.” Over a series of slides that include in-depth graphic representations and detailed analysis of real data, he shows how LashBack’s data insights can provide businesses with the clarity needed to understand a company’s own inboxing rates. For example, quality issues such as authentication, length of HTML, and use of special characters will impact delivery.
About LashBack®
Founded in 2003 by marketers with a vision of how to make email more safe and effective, LashBack® has grown to become an authority and go-to resource for email compliance and intelligence.
LashBack® is a patent-holder with a long history of innovation, collecting and leveraging data, and advocating for best practices. We continue to focus on creative ways to bring insights to our clients with services spanning every aspect of email marketing and a recent expansion into additional digital marketing channels. Whether you represent a large brand or are just getting started -- are more focused on risk or growth -- we have the information and services to help you accomplish your objectives.
DDoS In Oceania and the Pacific, presented by Dave Phelan at NZNOG 2024
LashBack Presentation at Mailcon January 2018
1. Data-driven Insights and Opportunities in Email
Peter Wilson - CEO, LashBack
Mailcon – Las Vegas
January 2018
2. January 2018 2
LashBack background/focus
• Based in St. Louis and providing services for more than a decade
• We have more than 100 major brands, agencies and networks as clients
• We provide critical visibility and information to businesses using email to acquire new customers
Manage compliance
Protect your brands and data
Identify risks and opportunities
PMA Compliance Council
3. January 2018 3
Quality email is powerful and achievable
• Email is a powerful marketing channel with an unmatched ROI
• The risks can be effectively managed
• There are critical insights that can be gained through data and analytics regarding offers, compliance,
quality, partners and delivery
• We currently offer services that provide compliance monitoring, list management, brand protection
and competitive insights
• In 2018, we are rolling out additional services focused on quality, partners and delivery
• Inboxing dramatically impacts the performance of an offer
• Message quality dramatically impacts inboxing
• Quality includes, but goes beyond, compliance – the presence of an unsubscribe and postal address
are compliance issues that impact delivery, but delivery is also impacted by broader quality issues like
authentication, length of HTML, and the use of special characters
4. January 2018 4
Discussion today
How do message characteristics/quality and partners impact delivery
• Passing authentication and the presence of a list-unsubscribe in the header and clear text unsubscribe
• Other characteristics that impact delivery, such as the number of links and images
• What is the impact of “spam text”... does it work
• What ESPs have the strongest delivery by email provider
• Some concluding thoughts
5. 5
LashBack data/disclaimer
September 2016
• LashBack collects millions of B2C email messages through a variety of sources. All of the email is
provided directly to LashBack or indirectly with the permission of the consumer
• LashBack does not market to consumers or sell personally identifiable information
All of the observations in this presentation are based on data received and analyzed by LashBack.
Presentations are intended for educational purposes only and do not replace independent professional
judgment. The observations may be influenced by multiple factors including the date, source, and size of the
data set and your results or experience may be meaningfully different. We do not make any representations
or offer any warranties about the completeness, reliability and accuracy of this information. We are strong
advocates for quality email marketing and do not advocate the use of any technique or practice in a vacuum
or that is counter to established best practices. Any action you take upon the information presented is strictly
at your own risk.
7. Some context on inboxing
Everyone has their own niche and varied experience, but what does the data say about:
• How does recent inboxing compare to inboxing a year ago?
• Is it getting better or worse?
• How does it vary by email provider?
January 2018 7
8. Inboxing similar to last year, but the rates and trends vary
Gmail at 60% (down 7%)
Yahoo at 50% (flat/up 1%)
Microsoft at 43% (up 4%)
January 2018 8
9. The impact of select quality characteristics
A few core quality measures significantly impact delivery – especially when all are present
• Passing authentication
• A list-unsubscribe in the header
• The presence of a clear text unsubscribe
How meaningful is the impact?
Does it vary by email provider?
January 2018 9
10. Passing versus failing authentication
January 2018 10
Average
difference:
22%
Compares messages that failed at least one authentication check to those that had no failures; messages without authentication were not considered.
11. Presence of a list-unsubscribe header
January 2018 11
Average
difference:
24%
Major impact
with Yahoo
12. Presence of a clear text unsubscribe
January 2018 12
Average
difference:
22%
Major impact
with Yahoo
Examples include “Click here to unsubscribe”, “To stop receiving these messages”, “to end your subscription”, etc.
13. Presence of a list-unsubscribe header & a clear text unsubscribe
January 2018 13
Average
difference:
33%
Yahoo
difference is
53%
14. Presence of all 3 – authentication, list-unsub & clear text unsub
January 2018 14
Average
difference:
38%
Yahoo
difference
is 55%
15. A few other important factors that influence delivery
Let’s take a look at the impact of:
• # of special characters in subject line
• Length of HTML portion of message
• # of links in HTML
• # of images
• Combination of HTML and text
January 2018 15
16. Negative impact of special characters in the subject line
January 2018 16
Number of characters
that were not A-Z, 0-9 or
whitespace
17. Positive correlation between inboxing and length of HTML
January 2018 17
By length of (number of
characters in) the HTML
portion of the message
18. Correlation between links in the HTML and inboxing
January 2018 18
Number of links in the
HTML portion of
messages.
Note: the dip near 155-
164 was caused by a
single sender sending a
high volume of phony
job offers
20. Positive impact of images on inboxing
January 2018 20
Number of images in the
HTML portion of
messages (1-25 images)
Interesting: for
Microsoft, zero images
seems to be better than
1-3
21. Varied impact of email format/Gmail
January 2018 21
Gmail does not
seem to
significantly
consider format
in determining
delivery
22. Microsoft seems to penalize text-only messages
January 2018 22
There is a 20%+
drop-off in the
inboxing of text-
only messages to
Microsoft
recipients
23. Yahoo appears to penalize HTML-only and favor text-only
January 2018 23
Yahoo inboxing
appears far
strongest for
text-only
messages.
24. What are the strongest influencers of inboxing?
We ran a regression against a large set of B2C email messages and data points within the messages to assess:
What combination of characteristics most strongly influences inboxing (across all
three major email providers)?
The result was…
Messages that include seven characteristics were 20%+ more likely to get to the inbox
across all three email providers
January 2018 24
25. The influence of the 7 characteristics on inboxing
January 2018 25
1. At least 10 images
2. Includes both HTML and text
parts
3. Length of HTML between
20,000 and 30,000
4. No more than 5 non-
alphanumeric characters
5. Contains a clear text
unsubscribe link
6. Contains a list-unsubscribe
header
7. No authentication failures
26. What characteristics are the most predictive
From a slightly different perspective –
• There are many data points and variables in a B2C email message, almost all of which influence delivery
• What are the 10 characteristics that enable you to best predict where a message will be
delivered?
• Based on our analysis… we can predict where the message will be received with 85% accuracy if we
know only these 10 characteristics
January 2018 26
Message characteristics are often asymmetrically correlated -- meaning that they do not equally predict
bulk and inbox. For example, the absence of a clear text unsubscribe strongly predicts that a message
will go to the bulk folder, but its presence is not a strong predictor that the message will be delivered to
the inbox.
The characteristics that are the most powerful (predictive) are at either end of the spectrum.
28. Our top 10… the most predictive list
January 2018 28
• Absence of a list-unsubscribe header (predicts bulk)
• Absence of a clear text unsubscribe (predicts bulk)
• Length of text within HTML part (predicts inbox)
• Total length of HTML part (predicts inbox)
• Total length of message body (predicts inbox)
• # of HTML tags (predicts inbox)
• # of unique HTML tags (predicts inbox)
• # of images (predicts inbox)
• # of linked images (predicts inbox)
• # of links in HTML (predicts inbox)
29. Number one among the top 10
January 2018 29
What was the one characteristic that was the most predictive of whether a message is delivered to the inbox?
The length of the HTML in the message was the most important factor in predicting delivery to the
inbox.
30. January 2018 30
Does “spam text” in a message work?
Reduces
delivery
16% to
Gmail
Reduces
delivery
27% to
Microsoft
Improves
delivery
18% to
Yahoo
We looked at the inboxing of all messages by CEP and compared it to the inboxing of messages manually flagged
in our system as containing spam text; the data is from the October/November 2017 time frame.
31. What ESPs have the strongest delivery across all three CEPs?
January 2018 31
Gmail avg 62% Yahoo avg 46%
Microsoft avg 41%
32. Varied delivery across the top three (volume) ESPs
January 2018 32
Gmail avg 62%
Yahoo avg 46%
Microsoft avg 41%
33. High volume ESPs with above average inboxing
Ten of the highest volume ESPs:
• AWeber
• BlueHornet
• Bronto
• Campaigner
• iContact
• InfusionSoft
• Maropost
• SendGrid
• SendLabs
• Yesmail
January 2018 33
Gmail Microsoft Yahoo
1 Yesmail SendGrid Yesmail
2 SendGrid Yesmail Campaigner
3 Bronto Maropost
4 Campaigner SendGrid
5 AWeber iContact
6 InfusionSoft
7 Bronto
8 SendLabs
9 BlueHornet
10 AWeber
Which of the ten had above average inboxing for each CEP:
Ranked by inboxing, November 2017
34. 34
Takeaways
January 2018
Email is a powerful but complex marketing channel. In order to succeed, you need to do the right
analytics… at the right time, using the right data
• In general, quality email gets better delivery
• On a more granular level, certain characteristics and combinations of characteristics seem to have
the greatest bearing on delivery (and the impact varies by email provider)
• Out of the many components of a message, the following seem to be highly impactful/predictive:
• No authentication failure
• Contains a list-unsubscribe header
• Contains a clear text unsubscribe link
• No or few non-alphanumeric characters in the subject line
• 10 or more links and images
• Includes both HTML and text parts
• Length of HTML between 20,000 and 30,000
Notes de l'éditeur
Delta:
Gmail: +23% (~5 > overall)
Microsoft: +15% (~5 > overall)
Yahoo: +29% (~9 > overall)
Compares messages that failed at least one authentication check (SPF, DKIM, etc.) to messages that had no failures.
Messages without any authentication (no SPF policy, no DKIM signature) were not considered.
Delta:
Gmail: +15% (~5 > overall)
Microsoft: +15% (~5 > overall)
Yahoo: +29% (~28 > overall)
Compares messages that contained a List-Unsubscribe header to ones that did not.
Delta:
Gmail: +16% (~5 > overall)
Microsoft: +13% (~4 > overall)
Yahoo: +36% (~16 > overall)
Compares messages that contained a reference to a clear text unsubscribe mechanism to ones that did not.
Examples:
“Click here to unsubscribe”, “To stop receiving these messages”, “to end your subscription”, etc.
Delta:
Gmail: +26 (~6 > overall)
Microsoft: +5 (~4 > overall)
Yahoo: +53 (~29 > overall)
Compares messages that contained both a List-Unsubscribe header and a clear text unsubscribe mechanism to messages that had neither.
Delta:
Gmail: +32 (~6 > overall)
Microsoft: +27 (~6 > overall)
Yahoo: +55 (~30 > overall)
Compares messages that pass authentication and contained a List-Unsubscribe header and a clear text unsubscribe mechanism to messages that failed authentication and had neither type of unsubscribe.
Percentage of messages delivered to the inbox
by number of characters that were not A-Z, 0-9 or whitespace.
Inbox delivery drops swiftly as the number of characters increases.
Percentage of messages delivered to the inbox
by length of (number of characters in) the HTML portion of messages.
Inboxing is significantly higher for longer messages.
***** ANIMATIONS!!! *****
Percentage of messages delivered to the inbox
Number of links in the HTML portion of messages. The orange area is % of total messages.
Inbox delivery suffers when there are few links in the message. Note that the vast majority of messages fall into the 1-29 range (orange area).
**** The dip near 155-164 was caused by a single sender blasting large lists of phony job openings. *****
Percentage of messages delivered to the inbox
Number of links in the HTML portion of messages. (1-25 links)
Percentage of messages delivered to the inbox
Number of images in the HTML portion of messages. (1-25 images)
Interestingly, for Microsoft, zero images seems to be better than 1-3.
Percentage of messages delivered to the inbox
By whether the message had only a text part, only an HTML part, or both.
Percentage of messages delivered to the inbox
By whether the message had only a text part, only an HTML part, or both.
Percentage of messages delivered to the inbox
By whether the message had only a text part, only an HTML part, or both.
Percentage of messages delivered to the inbox
By whether the message had only a text part, only an HTML part, or both.
Percentage of messages delivered to the inbox
By whether the message had only a text part, only an HTML part, or both.
Percentage of messages delivered to the inbox
By whether the message had only a text part, only an HTML part, or both.