An Inside Scoop on Origin's Email Marketing

From 2% to 25%: Improving Open Rates

GM! I’m Jon, and I work on data engineering projects at Origin Protocol. Email marketing isn’t something you’d think would fall under that role. Still, I’ve worked on it extensively over the years, so I was happy to help when we looked to revamp the processes and tools we used for our email marketing strategy.

Most of my previous work on email marketing was in the highly regulated financial services sector, which tends to err on the side of caution and is highly process-driven. Most of our issues were data-related, and building better processes helped our marketing team get back on track quickly. We improved our open rate from ~2% to over 25% and decreased our bounce rate from 5% to 0.01%.

DISCLAIMER: This will not be an in-the-weeds technical document but rather discuss some of our issues, solutions, and high-level approach to email marketing. This document will share some of our processes to help other teams improve their communication strategies or give our community greater visibility into our strategy.

The Problem(s)

We had some problems with email that are likely familiar to anyone with an extensive mailing list.

Most email solutions are expensive and overkill compared to the features used. We needed something simple to send a couple of emails out each month to our subscribers.
We had an older email list and no clear process to manage list hygiene. Our data quality could have been better, and there needed to be a process to improve this.
Facing deliverability issues due to high bounce rates and sporadic send volumes. We needed to improve our reputation and show that we weren’t mindlessly sending emails.
There was no isolation from our team emails, leading to additional risk when sending mass emails. Shared domain reputation can lead to flagging on your primary domain with some ISPs/ESPs.

Email marketing best practices are well documented online, so we’ll avoid talking about this and skip to discussing the tools and services we landed on when building our stack. From there, we will discuss list hygiene, deliverability, and how we handle email marketing at Origin.

Our Stack

Email validation: ZeroBounce

There are many email validation tools out there. I’ve used ZeroBounce in the past, so it was an obvious choice when we needed to validate our large database of emails. With API access and CSV upload, you can validate on the fly or use this to clean email lists.

Email marketing software billing is on the subscriber level and per email sent. Some plans allow for some quantity of each and then usually charge on some incremental level. You can think of the two components that make up email marketing software as subscriber management and email send servers. Said another way, you have to pay to store email addresses and to send emails to them. If you don’t have a clean subscriber list, you’re wasting money on subscriber storage and emails that bounce or are unread.

Subscriber management: Listmonk

We needed something lightweight and decoupled from the email server for our subscriber management and UI. We had minimal requirements, sending a few update emails per month (more on that below). Listmonk is an open-source tool that handles subscriber management and the email send interface. It offers some simple tools for bounce management, works well out of the box with SendGrid and SES, and provides some API access.

Email send server: Amazon SES

There are many third-party email send service options; you can even host one yourself with Postal. We opted not to deal with running our own email send server and chose to go with Amazon SES. We use AWS for some of our infrastructure at Origin, so SES offered what we were looking for. To provide some isolation from our team emails, we opted to use a subdomain as our send domain.

Rendering tests: Email on Acid

Handling rendering across email clients can be a pain in the ass. Listmonk gives us a lot of flexibility to build our base templates, and we lean into Email on Acid to ensure we’re not missing anything. Many email rendering issues come down to improperly inlined CSS (though not the only reason for poor rendering). You can use a tool like this one from Litmus to help with inlining CSS.

Deliverability tracking: Postmaster

Deliverability can be complicated, so we will cover it in more detail below. Some ESPs are more complicated to track than others (cough, Yahoo, cough), but it’s easier to just look at the data. Almost 75% of our email subscribers have Gmail addresses, with 95.6% of our engaged subscribers using Gmail. Albeit fairly rudimentary, we use Postmaster to keep a high-level view of email performance.

Blacklist monitoring: ZeroBounce

ZeroBounce also provides a free blacklist tool that checks several registries. ZeroBounce is one of many solutions out there, but it works.

These factors aren’t an exhaustive list of all the individual components in our email marketing stack, but it provides enough color for our approach. Additionally, we automate the syncing of different pieces of data, updating subscriber status, and various kinds of reporting.

List Hygiene

List hygiene is an essential process in ongoing email marketing efforts. Poor list hygiene leads to higher bounce rates, higher spam rates, lower deliverability, and overall reduces the effectiveness of your email marketing strategy.

A basic definition of email list hygiene is the process of ensuring that the subscribers you plan on sending to in the future are ready and willing to receive your emails.

Previously, we had no standardized process for managing email list hygiene. Recently, I had a chance to set this up properly.

Validating New Subscribers

Unless you have a hygiene process, I recommend scrubbing (validating) your list of subscriber emails. Assuming you have an ongoing hygiene strategy, you generally only need to rerun a validation process if you add large lists of new subscribers. The way this process works is that the provider handling the validation will run a series of tests on each email in your subscriber list.

A non-exhaustive list of these tests includes mailbox status, inbox quota, spam traps, abuse checks, and catchalls. The goal of these checks is to make sure you’re sending to an active inbox, your email isn’t going to bounce, the subscriber inbox is not full, and the subscriber quality is sufficient.

At Origin, we use ZeroBounce to validate large subscriber lists and for real-time email subscriber validation. Many other providers offer this service, and many email marketing tools can also handle this. Depending on your list size, this can run anywhere from $0.001 to $0.01 per email. It can get expensive for larger email list sizes, but it’s a necessary investment when introducing new emails.

You can upload a CSV of your subscriber list and receive some classification data for your list. After this, you can filter out emails that don’t meet your internal criteria. We remove catchalls, spam traps, abusers, dead mailboxes, and flag subscribers over their quota for reintroduction in the future.

Handling Bounces

The work doesn’t stop when you’re past the new subscriber stage. Bounces are a signal that gives you some insight into your email subscribers. Bounces come in two flavors: hard and soft bounces and a wide range of bounce statuses that can be specific to the recipient's email server.

Listmonk offers some tools to handle bounces automatically. Our standard process is to permanently remove any subscribers with a hard bounce and allow for 2-3 soft bounces before temporarily removing subscribers. Subscribers that soft bounce are periodically re-added to our send list and permanently removed if they receive a block a second time.

Subscriber Engagement

Subscriber engagement can be one of the more challenging things to execute regarding email marketing. It’s not because it’s difficult, but we tend to want to keep every single email we’ve ever collected and make sure we send it to the largest list possible.

At a certain point, you need to break up with your subscribers. Generally, we look at subscriber engagement over 3, 6 and 12 months to see what users have fallen off. When subscribers do not engage with us over a certain period, they get temporarily removed from our regular sends. This process is helpful to improve overall engagement rates and make sure we’re limiting our messages to the people interacting with our communication in some form. Access to your email inbox is a privilege we don’t take lightly. We’re always here when subscribers are ready to engage again.

Deliverability

When we reworked our strategy, we decided to start from scratch. We would isolate our marketing communication from our team domain, giving us the flexibility to move quickly.

I won’t cover the standard operating procedures (you can Google that) around email warmup other than to say, make sure you’re monitoring your metrics while ramping up send volumes slowly and steadily.

What to Measure

We monitor several different factors to assess the health and effectiveness of our email marketing. This list is not exhaustive, but at a high level, it outlines what we are most concerned about.

The following five factors are our primary areas of concern:

Spam rate: Most email send services provide data on spam rates. Not all ESPs report spam rates back to the service provider. You can review this for Gmail using Postmaster.

Target: <= 0.05%

Bounce rate: Like spam rate, this is usually reported by your send provider. It is generally available as part of many email marketing tools. Ideally, this will distinguish between hard and soft bounces.

Target: <= 0.05%

Open rate: We look at the unique open rate as this best reflects performance on a subscriber level. It helps us answer what percentage of our subscriber list looks at our emails.

Target: >= 21.5%

Unsubscribe rate: Unsubscribes is the normal course of business when sending emails. If an unsubscribe rate is too high, it signals a misalignment between the subscriber and the content. Taking long breaks between sends, low-quality content, and using questionable lists can all contribute to a higher-than-desired unsubscribe rate.

Target: <= 0.1%

Blacklists: How ISPs/ESPs use blacklists is not always clear, but you should aim to never end up on one. These third-party databases receive feedback from service providers and are looking for high spam rates or abnormal levels of send volume. The above factors influence these final two factors.

Target: 0

These final two factors are influenced by the above factors. If you monitor and measure your primary factors frequently, you don't need to monitor the following regularly.

Domain reputation: Outside of monitoring blacklists, we review our domain reputation weekly. Domain reputation is like an email marketing credit score, and it helps us react to changes if necessary.

Target: Medium/High (Postmaster)

IP reputation: Depending on your send volumes and overall strategy, you will use a shared or dedicated IP. In both cases, it’s worth monitoring your IP reputation. Due to the high occurrence of Gmail addresses, we check in on Postmaster weekly.