First_look_Data
As part of our ongoing mission to help everyone send better emails, we recently took a random (and anonymous) sample of 50,000,000 transactional emails sent through Sendwithus and began to investigate the data.

To start, we wanted to classify the purpose and intent of every email. We devised a hacky (and unimaginative) algorithm to sort each email into one of nine categories.

Transactional Email Categories

Category Example Keywords and Terms
welcome verify your email address, thanks for signing up
reminder reminder, you still haven’t, expiring
billing invoice, billing, payment, credit card
status_update confirmed, shipped, received, printing
promotional newsletter, redeem, sale, tips to help you, coupon
social invited, reply, followed, shared
password password reset
garbage internal reports and testing emails
unknown category unclear

After categorizing 50,000,000 emails and templates, we began to look at open and click rates across these categories.

Email click rates were quite varied and appeared to be heavily dependent on sender brand and email content – a great topic for further investigation.

Open rates, on the other hand, showed some very notable patterns.

Screenshot 2014-08-19 10.51.39

Here’s the same data, shown as relative open rates per category:

Email open rates by category

So What Does this Mean?

Chances are you’re sending at least one of these transactional emails to every single user who uses your product. And the data shows that these emails matter — users are opening and engaging with them regularly, much more than traditional marketing emails.

Welcome emails, in particular, get opened a lot. Even boring billing-related emails see significantly higher open rates than most other email types.

In short: these emails are important to your users.

We’re Just Getting Started

We’ve only begun to investigate our data set, and we’ve got a giant list of questions worth investigating in hopes of finding interesting or helpful insights for everyone.

Here are a few of our favorites:

  • Does putting dynamic data in the subject line make a difference on open rate?
  • What is the optimal length for a subject line?
  • What specific words, if any, correlate with a higher open rate?
  • If action is required, is the email more likely to be opened? Clicked?
  • What are senders with the highest open rates doing differently?

As we dig deeper, we hope to answer these questions and further our mission to make email better for everyone.

Posted in NewsTagged

4 thoughts on “Digging Into Data From 50M Transactional Emails

Leave a Reply

Your email address will not be published. Required fields are marked *