“Stop sending emails like a marketer. Start sending email like a personal assistant.”

I’ve communicated this line in a lot of my presentations on growth, but I haven’t talked about in depth the evolution on how to get there. There is a clear evolution that companies follow in terms of evolving their emails and notifications, from not sending them at all, to sending one-off blasts to their entire audience, to creating a lifecycle, to having a holistic personalized messaging platform. Having worked on these systems at my last two companies, I thought it would be beneficial to outline these transitions for people earlier along in their process.

Phase 0: We Hate Email

“We hate email, so we don’t send it to our customers.”

Almost every company I have worked with has communicated this at some point in its early stages, and it’s always wrong. Except for very specific groups, people don’t dislike email; they dislike bad email. The path to figuring what email your customers would like to receive is largely to ask what kind of value do they see in my product, and can I deliver that value in email form.

Phase 1: Mass Promotional Email + Personalized Order Notifications
If you are a transactional service, you have to send personalized order notifications, so that is where most companies start. At some point later, companies start sending mass emails to their broader audience about certain things like new features, discounts, etc. This effort will show improvements in key metrics, but it is very unsophisticated.

Promotions train users to wait for promotions to order, decreasing profitability. Also, with this approach, marketing is assuming a cadence for the customer instead of adapting to the customer’s cadence (and ultimately improving customer cadence to increase lifetime value).

If you are engineering constrained, there are some simple optimizations to this approach that will improve your performance:

• Develop additional emails intended to drive habit formation (instead of just timely purchase). Examples include trending items, item sales, new merchants added, recommended items
• If emails are successful, test them as push notifications as well
• Take existing confirmation emails, and add marketing messages to them (other things to buy, set up a re-order, most popular items, etc.)
• Send every non-transactional email and tweak to confirmation emails as an experiment with an enabled and control group to prove impact on lifetime value vs. unsubscribes/push permissions/app deletions

Phase II: Moving to Lifecycle Messaging
In Phase II, these additional emails and notifications that have been successful in Phase I start to form an automated program that consistently drives additional engagement from customers. In order to address messaging fatigue, these email and notification templates are managed to a frequency per month based on the team’s expected value of a good customer. This frequency is not based on data, but if everyone used the product properly, what would ideal frequency look like. The goal is to use messaging to remind people of the service and reinforce the habit. They are paired with personalized discount emails intended to drive new use cases and increase frequency.

These emails and notification templates are also managed against each other, so messaging does not get stale. For example, if you have three templates outside of confirmation emails, and you sent template 1 last week, you would attempt to send templates 2 and 3 before sending template 1 again. Also, each week, these emails have new subject lines to present them from looking like the same email as the previous weeks.

Phase III: Holistic, Personalized Messaging
As the Phase II approach flattens in terms of the additional impact it can drive, companies shift toward a more holistic, personalized system. This is a considerable investment, which we made at Pinterest. Essentially, product and engineering determine for each customer:

• The right content
• The right time to send it (day of week and time or day)
• The right amount (how many emails and pushes to send)
• The right channel (email, push, or both)

This requires a team to develop a log of every email/push sent to a subscriber, when it was sent, when it was opened, when it was clicked, what downstream engagement occurred from clicks, and which template it was. All emails and notifications are run through the same experiment dashboard as product changes to understand the impact on all key metrics. From this, it needs to determine:

• The best day(s) of week and time(s) to send messages to each user
• A prioritization of the templates to send based on historical click through rates and/or purchase rates
• How many messages per a generic time frame maximizes lifetime value of each user

This usually starts via a rules based approach, and eventually becomes powered by machine learning. If you lack enough historical data on a user to do this, for example new users, you group people who used to look like those users as a segment i.e. previous new users and look at the best performing approach for them. Email can no longer be considered marketing at this point. It is considered an extension of the core product.

The team also starts optimizing deliverability through choosing better message transfer agent partners and segmenting IP addresses for different templates to isolate issues. The team may also start investing in more advanced security measures like DMARC.

This is a considerable investment, which is why most companies only start building this once there are sending millions of emails a day with a lot of history from operating in the first two phases originally. At this point, companies know the value of email, and can justify the investment.

—

In my opinion, every company should end up at phase III at some point. The question is how long it takes to get there. This varies based on engineering constraints, scale, and how long it takes emails and notifications to flatten off in terms of additional engagement by the previous phases. Outsourcing this to a marketing technology company is also very problematic as it requires access to all of your user data, and any migration of data from system to system slows down performance. At a certain scale (like Pinterest), it is not even possible.

If you’re not at Pinterest’s level of sophistication, don’t dismay. Very few companies are. Just start to think about the long term evolution, and when is the right time to push for a step change in email and notification performance vs. continued optimization. It’s a big investment to shift from phase to phase, but the returns are usually worth it, and the impact of these emails and notifications in the current phase, and the struggle to improve their performance, should be what drives that decision to make additional investment to get to the next phase.

Currently listening to XTLP by μ-Ziq.